BUMP 1.6.2

fix sharding tests
SERVER-1607 more replica set checks before adding shard
2010-09-01 08:50:38 -04:00 · 2010-08-31 23:41:17 -04:00 · 2010-08-31 21:21:18 -04:00 · 2010-08-31 21:21:06 -04:00 · 2010-08-31 21:20:58 -04:00 · 2010-08-31 21:20:16 -04:00
81 changed files with 981 additions and 351 deletions
--- a/bson/bsoninlines.h
+++ b/bson/bsoninlines.h
@@ -387,16 +387,7 @@ namespace mongo {
            }
            break;
        case NumberDouble:
-            {
-                stringstream tmp;
-                tmp.precision( 16 );
-                tmp << number();
-                string n = tmp.str();
-                s << n;
-                // indicate this is a double:
-                if( strchr(n.c_str(), '.') == 0 && strchr(n.c_str(), 'E') == 0 && strchr(n.c_str(), 'N') == 0 )
-                    s << ".0";
-            }
+            s.appendDoubleNice( number() );
            break;
        case NumberLong:
            s << _numberLong();
--- a/bson/util/builder.h
+++ b/bson/util/builder.h
@@ -164,6 +164,7 @@ namespace mongo {
 #define SBNUM(val,maxSize,macro) \
            int prev = _buf.l; \
            int z = sprintf( _buf.grow(maxSize) , macro , (val) );  \
+            assert( z >= 0 ); \
            _buf.l = prev + z; \
            return *this; 

@@ -197,6 +198,17 @@ namespace mongo {
        }
 #undef SBNUM

+        void appendDoubleNice( double x ){
+            int prev = _buf.l;
+            char * start = _buf.grow( 32 );
+            int z = sprintf( start , "%.16g" , x );
+            assert( z >= 0 );
+            _buf.l = prev + z;
+            if( strchr(start, '.') == 0 && strchr(start, 'E') == 0 && strchr(start, 'N') == 0 ){
+                write( ".0" , 2 );
+            }
+        }
+
        void write( const char* buf, int len){
            memcpy( _buf.grow( len ) , buf , len );
        }
--- a/client/dbclient.cpp
+++ b/client/dbclient.cpp
@@ -1063,6 +1063,18 @@ namespace mongo {
        return checkMaster()->findOne(a,b,c,d);
    }

+    bool DBClientReplicaSet::isMember( const DBConnector * conn ) const {
+        if ( conn == this )
+            return true;
+        
+        for ( unsigned i=0; i<_conns.size(); i++ )
+            if ( _conns[i]->isMember( conn ) )
+                return true;
+        
+        return false;
+    }
+    
+
    bool serverAlive( const string &uri ) {
        DBClientConnection c( false, 0, 20 ); // potentially the connection to server could fail while we're checking if it's alive - so use timeouts
        string err;
--- a/client/dbclient.h
+++ b/client/dbclient.h
@@ -106,10 +106,11 @@ namespace mongo {
            _finishInit();
        }

-        ConnectionString( ConnectionType type , const vector<HostAndPort>& servers )
-            : _type( type ) , _servers( servers ){
-            _finishInit();
-        }
+        // TODO Delete if nobody is using
+        //ConnectionString( ConnectionType type , const vector<HostAndPort>& servers )
+        //    : _type( type ) , _servers( servers ){
+        //    _finishInit();
+        //}
        
        ConnectionString( ConnectionType type , const string& s , const string& setName = "" ){
            _type = type;
@@ -156,6 +157,14 @@ namespace mongo {

        static ConnectionString parse( const string& url , string& errmsg );
        
+        string getSetName() const{
+            return _setName;
+        }
+
+        vector<HostAndPort> getServers() const {
+            return _servers;
+        }
+        
    private:

        ConnectionString(){
@@ -314,7 +323,7 @@ namespace mongo {
 /** Typically one uses the QUERY(...) macro to construct a Query object.
    Example: QUERY( "age" << 33 << "school" << "UCLA" )
 */
-#define QUERY(x) Query( BSON(x) )
+#define QUERY(x) mongo::Query( BSON(x) )

    /**
       interface that handles communication with the db
@@ -329,6 +338,8 @@ namespace mongo {

        /* used by QueryOption_Exhaust.  To use that your subclass must implement this. */
        virtual void recv( Message& m ) { assert(false); }
+
+        virtual string getServerAddress() const = 0;
    };

    /**
@@ -739,8 +750,6 @@ namespace mongo {
         */
        virtual void update( const string &ns , Query query , BSONObj obj , bool upsert = false , bool multi = false );
        
-        virtual string getServerAddress() const = 0;
-        
        virtual bool isFailed() const = 0;
        
        virtual void killCursor( long long cursorID ) = 0;
@@ -758,6 +767,9 @@ namespace mongo {
        virtual void say( Message& toSend  ) = 0;

        virtual ConnectionString::ConnectionType type() const = 0;
+
+        /** @return true if conn is either equal to or contained in this connection */
+        virtual bool isMember( const DBConnector * conn ) const = 0;
    }; // DBClientBase
    
    class DBClientReplicaSet;
@@ -892,11 +904,16 @@ namespace mongo {
        virtual bool call( Message &toSend, Message &response, bool assertOk = true );
        
        virtual ConnectionString::ConnectionType type() const { return ConnectionString::MASTER; }  
+
+        virtual bool isMember( const DBConnector * conn ) const { return this == conn; };
+
+        virtual void checkResponse( const char *data, int nReturned );
+
    protected:
        friend class SyncClusterConnection;
        virtual void recv( Message& m );
        virtual void sayPiggyBack( Message &toSend );
-        virtual void checkResponse( const char *data, int nReturned );
+
    };
    
    /** Use this class to connect to a replica set of servers.  The class will manage
@@ -987,9 +1004,12 @@ namespace mongo {

        virtual ConnectionString::ConnectionType type() const { return ConnectionString::SET; }  

+        virtual bool isMember( const DBConnector * conn ) const;
+
+        virtual void checkResponse( const char *data, int nReturned ) { checkMaster()->checkResponse( data , nReturned ); }
+
    protected:                
-        virtual void sayPiggyBack( Message &toSend ) { assert(false); }
-        virtual void checkResponse( const char *data, int nReturned ) { assert(false); }
+        virtual void sayPiggyBack( Message &toSend ) { checkMaster()->say( toSend ); }
        
        bool isFailed() const {
            return _currentMaster == 0 || _currentMaster->isFailed();
--- a/client/dbclientcursor.cpp
+++ b/client/dbclientcursor.cpp
@@ -193,7 +193,7 @@ namespace mongo {

    void DBClientCursor::attach( AScopedConnection * conn ){
        assert( _scopedHost.size() == 0 );
-        assert( connector == conn->get() );
+        assert( conn->get()->isMember( connector ) );
        _scopedHost = conn->getHost();
        conn->done();
        connector = 0;
--- a/client/distlock.cpp
+++ b/client/distlock.cpp
@@ -109,9 +109,6 @@ namespace mongo {

       
    bool DistributedLock::lock_try( string why , BSONObj * other ){
-        // check for recrusive
-        assert( getState() == 0 );
-
        ScopedDbConnection conn( _conn );
            
        BSONObjBuilder queryBuilder;
@@ -136,7 +133,7 @@ namespace mongo {
                    return false;
                }

-                Date_t elapsed = jsTime() - lastPing["ping"].Date(); // in ms
+                unsigned long long elapsed = jsTime() - lastPing["ping"].Date(); // in ms
                elapsed = elapsed / ( 1000 * 60 ); // convert to minutes

                if ( elapsed <= _takeoverMinutes ){
@@ -208,18 +205,33 @@ namespace mongo {
        if ( ! gotLock )
            return false;
            
-        _state.set( 1 );
        return true;
    }

    void DistributedLock::unlock(){
-        ScopedDbConnection conn( _conn );
-        conn->update( _ns , _id, BSON( "$set" << BSON( "state" << 0 ) ) );
-        log(1) << "dist_lock unlock: " << conn->findOne( _ns , _id ) << endl;
-        conn.done();
+        const int maxAttempts = 3;
+        int attempted = 0;
+        while ( ++attempted <= maxAttempts ) {
+
+            try {
+                ScopedDbConnection conn( _conn );
+                conn->update( _ns , _id, BSON( "$set" << BSON( "state" << 0 ) ) );
+                log(1) << "dist_lock unlock: " << conn->findOne( _ns , _id ) << endl;
+                conn.done();
+
+                return;
+
            
-        _state.set( 0 );
+            } catch ( std::exception& e) {
+                log( LL_WARNING ) << "dist_lock  " << _name << " failed to contact config server in unlock attempt " 
+                                  << attempted << ": " << e.what() <<  endl;
+
+                sleepsecs(1 << attempted);
+            }
+        }
+
+        log( LL_WARNING ) << "dist_lock couldn't consumate unlock request. " << "Lock " << _name 
+                              << " will be taken over after " <<  _takeoverMinutes << " minutes timeout" << endl;
    }

-
 }
--- a/client/distlock.h
+++ b/client/distlock.h
@@ -36,14 +36,6 @@ namespace mongo {
         */
        DistributedLock( const ConnectionString& conn , const string& name , unsigned takeoverMinutes = 10 );

-        int getState(){
-            return _state.get();
-        }
-
-        bool isLocked(){
-            return _state.get() != 0;
-        }
-        
        bool lock_try( string why , BSONObj * other = 0 );
        void unlock();

@@ -54,8 +46,6 @@ namespace mongo {
        
        string _ns;
        BSONObj _id;
-        
-        ThreadLocalValue<int> _state;
    };
    
    class dist_lock_try {
--- a/client/gridfs.cpp
+++ b/client/gridfs.cpp
@@ -66,7 +66,6 @@ namespace mongo {
    }

    BSONObj GridFS::storeFile( const char* data , size_t length , const string& remoteName , const string& contentType){
-        massert( 10279 , "large files not yet implemented", length <= 0xffffffff);
        char const * const end = data + length;

        OID id;
@@ -127,8 +126,6 @@ namespace mongo {
        if (fd != stdin)
            fclose( fd );
        
-        massert( 10280 , "large files not yet implemented", length <= 0xffffffff);
-
        return insertFile((remoteName.empty() ? fileName : remoteName), id, length, contentType);
    }

--- a/client/parallel.cpp
+++ b/client/parallel.cpp
@@ -328,6 +328,38 @@ namespace mongo {
    void ParallelSortClusteredCursor::_finishCons(){
        _numServers = _servers.size();
        _cursors = 0;
+
+        if ( ! _sortKey.isEmpty() && ! _fields.isEmpty() ){
+            // we need to make sure the sort key is in the project
+            bool isNegative = false;
+            BSONObjBuilder b;
+            {
+                BSONObjIterator i( _fields );
+                while ( i.more() ){
+                    BSONElement e = i.next();
+                    b.append( e );
+                    if ( ! e.trueValue() )
+                        isNegative = true;
+                }
+            }                    
+            
+            {
+                BSONObjIterator i( _sortKey );
+                while ( i.more() ){
+                    BSONElement e = i.next();
+                    BSONElement f = _fields.getField( e.fieldName() );
+                    if ( isNegative ){
+                        uassert( 13431 , "have to have sort key in projection and removing it" , f.eoo() );
+                    }
+                    else if ( f.eoo() ){
+                        // add to projection
+                        b.append( e );
+                    }
+                }
+            }
+            
+            _fields = b.obj();
+        }
    }
    
    void ParallelSortClusteredCursor::_init(){
--- a/client/syncclusterconnection.cpp
+++ b/client/syncclusterconnection.cpp
@@ -369,4 +369,16 @@ namespace mongo {
        // should never need to do this
        assert(0);
    }
+
+    bool SyncClusterConnection::isMember( const DBConnector * conn ) const {
+        if ( conn == this )
+            return true;
+        
+        for ( unsigned i=0; i<_conns.size(); i++ )
+            if ( _conns[i]->isMember( conn ) )
+                return true;
+        
+        return false;
+    }
+
 }
--- a/client/syncclusterconnection.h
+++ b/client/syncclusterconnection.h
@@ -90,6 +90,8 @@ namespace mongo {

        virtual ConnectionString::ConnectionType type() const { return ConnectionString::SYNC; }  

+        virtual bool isMember( const DBConnector * conn ) const;
+
    private:
        SyncClusterConnection( SyncClusterConnection& prev );
        string _toString() const;        
--- a/db/client.cpp
+++ b/db/client.cpp
@@ -259,6 +259,29 @@ namespace mongo {
        }
    }

+    BSONObj CurOp::query( bool threadSafe ) {
+        if( querySize() == 1 ) { 
+            return _tooBig;
+        }
+        
+        if ( ! threadSafe ){
+            BSONObj o(_queryBuf);
+            return o;
+        }
+
+        int size = querySize();        
+        int before = checksum( _queryBuf , size );
+        BSONObj a(_queryBuf);
+        BSONObj b = a.copy();
+        int after = checksum( _queryBuf , size );
+        
+        if ( before == after )
+            return b;
+        
+        return BSON( "msg" << "query changed while capturing" );
+    }
+
+
    BSONObj CurOp::infoNoauth( int attempt ) {
        BSONObjBuilder b;
        b.append("opid", _opNum);
@@ -402,7 +425,7 @@ namespace mongo {
                    tablecell( ss , co.getOp() );
                    tablecell( ss , co.getNS() );
                    if ( co.haveQuery() )
-                        tablecell( ss , co.query() );
+                        tablecell( ss , co.query( true ) );
                    else
                        tablecell( ss , "" );
                    tablecell( ss , co.getRemoteString() );
--- a/db/clientcursor.cpp
+++ b/db/clientcursor.cpp
@@ -32,27 +32,37 @@

 namespace mongo {

-    CCById ClientCursor::clientCursorsById;
-    CCByLoc ClientCursor::byLoc;
-    boost::recursive_mutex ClientCursor::ccmutex;
+    typedef multimap<DiskLoc, ClientCursor*> CCByLoc;

-    unsigned ClientCursor::byLocSize() { 
+    CCById ClientCursor::clientCursorsById;
+    boost::recursive_mutex ClientCursor::ccmutex;
+    long long ClientCursor::numberTimedOut = 0;
+
+    /*static*/ void ClientCursor::assertNoCursors() { 
        recursive_scoped_lock lock(ccmutex);
-        return byLoc.size();
+        if( clientCursorsById.size() ) { 
+            log() << "ERROR clientcursors exist but should not at this point" << endl;
+            ClientCursor *cc = clientCursorsById.begin()->second;
+            log() << "first one: " << cc->cursorid << ' ' << cc->ns << endl;
+            clientCursorsById.clear();
+            assert(false);
+        }
    }

+
    void ClientCursor::setLastLoc_inlock(DiskLoc L) {
        if ( L == _lastLoc )
            return;

+        CCByLoc& bl = byLoc();
        if ( !_lastLoc.isNull() ) {
-            CCByLoc::iterator i = kv_find(byLoc, _lastLoc, this);
-            if ( i != byLoc.end() )
-                byLoc.erase(i);
+            CCByLoc::iterator i = kv_find(bl, _lastLoc, this);
+            if ( i != bl.end() )
+                bl.erase(i);
        }

        if ( !L.isNull() )
-            byLoc.insert( make_pair(L, this) );
+            bl.insert( make_pair(L, this) );
        _lastLoc = L;
    }

@@ -76,24 +86,52 @@ namespace mongo {
        {
            recursive_scoped_lock lock(ccmutex);

-            for ( CCByLoc::iterator i = byLoc.begin(); i != byLoc.end(); ++i ) {
+            Database *db = cc().database();
+            assert(db);
+            assert( str::startsWith(nsPrefix, db->name) );
+
+            for( CCById::iterator i = clientCursorsById.begin(); i != clientCursorsById.end(); ++i ) {
                ClientCursor *cc = i->second;
-                if ( strncmp(nsPrefix, cc->ns.c_str(), len) == 0 )
+                if( cc->_db != db ) 
+                    continue;
+                if ( strncmp(nsPrefix, cc->ns.c_str(), len) == 0 ) {
                    toDelete.push_back(i->second);
+                }
            }

+            /*
+            note : we can't iterate byloc because clientcursors may exist with a loc of null in which case
+                   they are not in the map.  perhaps they should not exist though in the future?  something to 
+                   change???
+                   
+            CCByLoc& bl = db->ccByLoc;
+            for ( CCByLoc::iterator i = bl.begin(); i != bl.end(); ++i ) {
+                ClientCursor *cc = i->second;
+                if ( strncmp(nsPrefix, cc->ns.c_str(), len) == 0 ) {
+                    assert( cc->_db == db );
+                    toDelete.push_back(i->second);
+                }
+            }*/
+
            for ( vector<ClientCursor*>::iterator i = toDelete.begin(); i != toDelete.end(); ++i )
                delete (*i);
        }
    }

+    bool ClientCursor::shouldTimeout( unsigned millis ){
+        _idleAgeMillis += millis;
+        return _idleAgeMillis > 600000 && _pinValue == 0;
+    }
+
    /* called every 4 seconds.  millis is amount of idle time passed since the last call -- could be zero */
    void ClientCursor::idleTimeReport(unsigned millis) {
+        readlock lk("");
        recursive_scoped_lock lock(ccmutex);
-        for ( CCByLoc::iterator i = byLoc.begin(); i != byLoc.end();  ) {
-            CCByLoc::iterator j = i;
+        for ( CCById::iterator i = clientCursorsById.begin(); i != clientCursorsById.end();  ) {
+            CCById::iterator j = i;
            i++;
            if( j->second->shouldTimeout( millis ) ){
+                numberTimedOut++;
                log(1) << "killing old cursor " << j->second->cursorid << ' ' << j->second->ns 
                       << " idle:" << j->second->idleTime() << "ms\n";
                delete j->second;
@@ -106,10 +144,12 @@ namespace mongo {
    */
    void ClientCursor::informAboutToDeleteBucket(const DiskLoc& b) {
        recursive_scoped_lock lock(ccmutex);
-        RARELY if ( byLoc.size() > 70 ) {
-            log() << "perf warning: byLoc.size=" << byLoc.size() << " in aboutToDeleteBucket\n";
+        Database *db = cc().database();
+        CCByLoc& bl = db->ccByLoc;
+        RARELY if ( bl.size() > 70 ) {
+            log() << "perf warning: byLoc.size=" << bl.size() << " in aboutToDeleteBucket\n";
        }
-        for ( CCByLoc::iterator i = byLoc.begin(); i != byLoc.end(); i++ )
+        for ( CCByLoc::iterator i = bl.begin(); i != bl.end(); i++ )
            i->second->c->aboutToDeleteBucket(b);
    }
    void aboutToDeleteBucket(const DiskLoc& b) {
@@ -120,8 +160,11 @@ namespace mongo {
    void ClientCursor::aboutToDelete(const DiskLoc& dl) {
        recursive_scoped_lock lock(ccmutex);

-        CCByLoc::iterator j = byLoc.lower_bound(dl);
-        CCByLoc::iterator stop = byLoc.upper_bound(dl);
+        Database *db = cc().database();
+        assert(db);
+        CCByLoc& bl = db->ccByLoc;
+        CCByLoc::iterator j = bl.lower_bound(dl);
+        CCByLoc::iterator stop = bl.upper_bound(dl);
        if ( j == stop )
            return;

@@ -139,6 +182,7 @@ namespace mongo {
        
        for ( vector<ClientCursor*>::iterator i = toAdvance.begin(); i != toAdvance.end(); ++i ){
            ClientCursor* cc = *i;
+            wassert(cc->_db == db);
            
            if ( cc->_doingDeletes ) continue;

@@ -157,7 +201,9 @@ namespace mongo {
            c->advance();
            if ( c->eof() ) {
                // advanced to end
-                // leave ClieneCursor in place so next getMore doesn't fail
+                // leave ClientCursor in place so next getMore doesn't fail
+                // still need to mark new location though
+                cc->updateLocation();
            }
            else {
                wassert( c->refLoc() != dl );
@@ -296,6 +342,13 @@ namespace mongo {

    int ctmLast = 0; // so we don't have to do find() which is a little slow very often.
    long long ClientCursor::allocCursorId_inlock() {
+        if( 0 ) { 
+            static long long z;
+            ++z;
+            cout << "TEMP alloccursorid " << z << endl;
+            return z;
+        }
+
        long long x;
        int ctm = (int) curTimeMillis();
        while ( 1 ) {
@@ -328,7 +381,13 @@ namespace mongo {
    }


-
+    void ClientCursor::appendStats( BSONObjBuilder& result ){
+        recursive_scoped_lock lock(ccmutex);
+        result.appendNumber("totalOpen", (int)clientCursorsById.size() );
+        result.appendNumber("clientCursors_size", (int) numCursors());
+        result.appendNumber("timedOut" , (int)numberTimedOut);
+    }
+    
    // QUESTION: Restrict to the namespace from which this command was issued?
    // Alternatively, make this command admin-only?
    class CmdCursorInfo : public Command {
@@ -339,11 +398,8 @@ namespace mongo {
            help << " example: { cursorInfo : 1 }";
        }
        virtual LockType locktype() const { return NONE; }
-        bool run(const string&, BSONObj& jsobj, string& errmsg, BSONObjBuilder& result, bool fromRepl ){
-            recursive_scoped_lock lock(ClientCursor::ccmutex);
-            result.append("totalOpen", unsigned( ClientCursor::clientCursorsById.size() ) );
-            result.append("byLocation_size", unsigned( ClientCursor::byLoc.size() ) );
-            result.append("clientCursors_size", unsigned( ClientCursor::clientCursorsById.size() ) );
+        bool run(const string& dbname, BSONObj& jsobj, string& errmsg, BSONObjBuilder& result, bool fromRepl ){
+            ClientCursor::appendStats( result );
            return true;
        }
    } cmdCursorInfo;
--- a/db/clientcursor.h
+++ b/db/clientcursor.h
@@ -45,8 +45,6 @@ namespace mongo {
    */
    typedef map<CursorId, ClientCursor*> CCById;

-    typedef multimap<DiskLoc, ClientCursor*> CCByLoc;
-
    extern BSONObj id_obj;

    class ClientCursor {
@@ -64,14 +62,13 @@ namespace mongo {
        ElapsedTracker _yieldSometimesTracker;

        static CCById clientCursorsById;
-        static CCByLoc byLoc;
-        static boost::recursive_mutex ccmutex;   // must use this for all statics above!
-        
-        static CursorId allocCursorId_inlock();
-
-
+        static long long numberTimedOut;
+        static boost::recursive_mutex ccmutex;   // must use this for all statics above!        
+        static CursorId allocCursorId_inlock();        
        
    public:
+        static void assertNoCursors();
+
        /* use this to assure we don't in the background time out cursor while it is under use.
           if you are using noTimeout() already, there is no risk anyway.
           Further, this mechanism guards against two getMore requests on the same cursor executing
@@ -139,19 +136,23 @@ namespace mongo {
        };

        /*const*/ CursorId cursorid;
-        string ns;
-        shared_ptr<Cursor> c;
+        const string ns;
+        const shared_ptr<Cursor> c;
        int pos;                  // # objects into the cursor so far 
        BSONObj query;
-        int _queryOptions;        // see enum QueryOptions dbclient.h
+        const int _queryOptions;        // see enum QueryOptions dbclient.h
        OpTime _slaveReadTill;
+        Database * const _db;

        ClientCursor(int queryOptions, shared_ptr<Cursor>& _c, const string& _ns) :
            _idleAgeMillis(0), _pinValue(0), 
            _doingDeletes(false), _yieldSometimesTracker(128,10),
            ns(_ns), c(_c), 
-            pos(0), _queryOptions(queryOptions)
+            pos(0), _queryOptions(queryOptions), 
+            _db( cc().database() )
        {
+            assert( _db );
+            assert( str::startsWith(_ns, _db->name) );
            if( queryOptions & QueryOption_NoCursorTimeout )
                noTimeout();
            recursive_scoped_lock lock(ccmutex);
@@ -308,10 +309,7 @@ namespace mongo {
        /**
         * @param millis amount of idle passed time since last call
         */
-        bool shouldTimeout( unsigned millis ){
-            _idleAgeMillis += millis;
-            return _idleAgeMillis > 600000 && _pinValue == 0;
-        }
+        bool shouldTimeout( unsigned millis );

        void storeOpForSlave( DiskLoc last );
        void updateSlaveLocation( CurOp& curop );
@@ -327,12 +325,18 @@ private:
        void noTimeout() { 
            _pinValue++;
        }
+
+        multimap<DiskLoc, ClientCursor*>& byLoc() { 
+            return _db->ccByLoc;
+        }
 public:
        void setDoingDeletes( bool doingDeletes ){
            _doingDeletes = doingDeletes;
        }
+        
+        static void appendStats( BSONObjBuilder& result );

-        static unsigned byLocSize();        // just for diagnostics
+        static unsigned numCursors() { return clientCursorsById.size(); }

        static void informAboutToDeleteBucket(const DiskLoc& b);
        static void aboutToDelete(const DiskLoc& dl);
--- a/db/cmdline.cpp
+++ b/db/cmdline.cpp
@@ -45,7 +45,7 @@ namespace mongo {
            ("bind_ip", po::value<string>(&cmdLine.bind_ip), "comma separated list of ip addresses to listen on - all local ips by default")
            ("logpath", po::value<string>() , "file to send all output to instead of stdout" )
            ("logappend" , "append to logpath instead of over-writing" )
-            ("pidfilepath", po::value<string>(), "directory for pidfile (if not set, no pidfile is created)")
+            ("pidfilepath", po::value<string>(), "full path to pidfile (if not set, no pidfile is created)")
 #ifndef _WIN32
            ("fork" , "fork server process" )
 #endif
--- a/db/curop.h
+++ b/db/curop.h
@@ -85,13 +85,7 @@ namespace mongo {
        int querySize() const { return *((int *) _queryBuf); }
        bool haveQuery() const { return querySize() != 0; }

-        BSONObj query() {
-            if( querySize() == 1 ) { 
-                return _tooBig;
-            }
-            BSONObj o(_queryBuf);
-            return o;
-        }
+        BSONObj query( bool threadSafe = false);

        void ensureStarted(){
            if ( _start == 0 )
--- a/db/database.h
+++ b/db/database.h
@@ -22,6 +22,8 @@

 namespace mongo {

+    class ClientCursor;
+
    /**
     * Database represents a database database
     * Each database database has its own set of files -- dbname.ns, dbname.0, dbname.1, ...
@@ -196,6 +198,9 @@ namespace mongo {
        NamespaceIndex namespaceIndex;
        int profile; // 0=off.
        string profileName; // "alleyinsider.system.profile"
+
+        multimap<DiskLoc, ClientCursor*> ccByLoc;
+
        int magic; // used for making sure the object is still loaded in memory 
    };

--- a/db/dbcommands.cpp
+++ b/db/dbcommands.cpp
@@ -126,6 +126,7 @@ namespace mongo {
                    assert( sprintf( buf , "w block pass: %lld" , ++passes ) < 30 );
                    c.curop()->setMessage( buf );
                    sleepmillis(1);
+                    killCurrentOp.checkForInterrupt();
                }
                result.appendNumber( "wtime" , t.millis() );
            }
@@ -159,32 +160,6 @@ namespace mongo {
        }
    } cmdGetPrevError;

-    class CmdSwitchToClientErrors : public Command {
-    public:
-        virtual bool requiresAuth() { return false; }
-        virtual bool logTheOp() {
-            return false;
-        }
-        virtual void help( stringstream& help ) const {
-            help << "convert to id based errors rather than connection based";
-        }
-        virtual bool slaveOk() const {
-            return true;
-        }
-        virtual LockType locktype() const { return NONE; } 
-        CmdSwitchToClientErrors() : Command("switchToClientErrors", false, "switchtoclienterrors") {}
-        bool run(const string& dbnamne , BSONObj& cmdObj, string& errmsg, BSONObjBuilder& result, bool fromRepl) {
-            if ( lastError.getID() ){
-                errmsg = "already in client id mode";
-                return false;
-            }
-            LastError *le = lastError.disableForCommand();
-            le->overridenById = true;
-            result << "ok" << 1;
-            return true;
-        }
-    } cmdSwitchToClientErrors;
-
    class CmdDropDatabase : public Command {
    public:
        virtual bool logTheOp() {
@@ -292,7 +267,6 @@ namespace mongo {
        }

        bool run(const string& dbname, BSONObj& cmdObj, string& errmsg, BSONObjBuilder& result, bool fromRepl) {
-            
            long long start = Listener::getElapsedTimeMillis();
            BSONObjBuilder timeBuilder(128);

@@ -381,7 +355,13 @@ namespace mongo {
                globalFlushCounters.append( bb );
                bb.done();
            }
-
+            
+            {
+                BSONObjBuilder bb( result.subobjStart( "cursros" ) );
+                ClientCursor::appendStats( bb );
+                bb.done();
+            }
+            
            timeBuilder.appendNumber( "after counters" , Listener::getElapsedTimeMillis() - start );            

            if ( anyReplEnabled() ){
@@ -648,10 +628,10 @@ namespace mongo {
        virtual void help( stringstream& help ) const {
            help << "create a collection";
        }
-        virtual bool run(const string& dbname , BSONObj& cmdObj, string& errmsg, BSONObjBuilder& result, bool) {
+        virtual bool run(const string& dbname , BSONObj& cmdObj, string& errmsg, BSONObjBuilder& result, bool fromRepl ) {
            string ns = dbname + '.' + cmdObj.firstElement().valuestr();
            string err;
-            bool ok = userCreateNS(ns.c_str(), cmdObj, err, true);
+            bool ok = userCreateNS(ns.c_str(), cmdObj, err, ! fromRepl );
            if ( !ok && !err.empty() )
                errmsg = err;
            return ok;
@@ -1450,17 +1430,18 @@ namespace mongo {
            
            BSONElementSet values;
            shared_ptr<Cursor> cursor = bestGuessCursor(ns.c_str() , query , BSONObj() );
+            scoped_ptr<ClientCursor> cc (new ClientCursor(QueryOption_NoCursorTimeout, cursor, ns));

            while ( cursor->ok() ){
-                if ( cursor->matcher() && ! cursor->matcher()->matchesCurrent( cursor.get() ) ){
-                    cursor->advance();
-                    continue;
+                if ( !cursor->matcher() || cursor->matcher()->matchesCurrent( cursor.get() ) ){
+                    BSONObj o = cursor->current();
+                    o.getFieldsDotted( key, values );
                }

-                BSONObj o = cursor->current();
                cursor->advance();
-                
-                o.getFieldsDotted( key.c_str(), values );
+
+                if (!cc->yieldSometimes())
+                    break;
            }

            BSONArrayBuilder b( result.subarrayStart( "values" ) );
--- a/db/dbcommands_generic.cpp
+++ b/db/dbcommands_generic.cpp
@@ -60,6 +60,13 @@ namespace mongo {
        bool run(const string& dbname, BSONObj& jsobj, string& errmsg, BSONObjBuilder& result, bool fromRepl ){
            result << "version" << versionString << "gitVersion" << gitVersion() << "sysInfo" << sysInfo();
            result << "bits" << ( sizeof( int* ) == 4 ? 32 : 64 );
+            result.appendBool( "debug" , 
+#ifdef _DEBUG
+                               true
+#else
+                               false
+#endif
+                               );
            return true;
        }
    } cmdBuildInfo;
--- a/db/dbwebserver.cpp
+++ b/db/dbwebserver.cpp
@@ -432,7 +432,7 @@ namespace mongo {
            ss << p( a("/", "back", "Home") );
            ss << p( "<b>MongoDB List of <a href=\"http://www.mongodb.org/display/DOCS/Commands\">Commands</a></b>\n" );
            const map<string, Command*> *m = Command::commandsByBestName();
-            ss << "S:slave-only  N:no-lock  R:read-lock  W:write-lock  A:admin-only<br>\n";
+            ss << "S:slave-ok  R:read-lock  W:write-lock  A:admin-only<br>\n";
            ss << table();
            ss << "<tr><th>Command</th><th>Attributes</th><th>Help</th></tr>\n";
            for( map<string, Command*>::const_iterator i = m->begin(); i != m->end(); i++ ) 
--- a/db/instance.h
+++ b/db/instance.h
@@ -143,6 +143,7 @@ namespace mongo {
        }

        virtual ConnectionString::ConnectionType type() const { return ConnectionString::MASTER; }  
+        virtual bool isMember( const DBConnector * conn ) const { return this == conn; };
    };

    extern int lockFile;
--- a/db/lasterror.cpp
+++ b/db/lasterror.cpp
@@ -168,9 +168,7 @@ namespace mongo {
    }
    
    LastError * LastErrorHolder::startRequest( Message& m , int clientId ) {
-
-        if ( clientId == 0 )
-            clientId = m.header()->id & 0xFFFF0000;
+        assert( clientId );
        setID( clientId );

        LastError * le = _get( true );
@@ -179,11 +177,7 @@ namespace mongo {
    }

    void LastErrorHolder::startRequest( Message& m , LastError * connectionOwned ) {
-        if ( !connectionOwned->overridenById ) {
-            prepareErrForNewRequest( m, connectionOwned );
-            return;
-        }
-        startRequest(m);
+        prepareErrForNewRequest( m, connectionOwned );
    }

    void LastErrorHolder::disconnect( int clientId ){
--- a/db/lasterror.h
+++ b/db/lasterror.h
@@ -32,7 +32,6 @@ namespace mongo {
        long long nObjects;
        int nPrev;
        bool valid;
-        bool overridenById;
        bool disabled;
        void writeback( OID& oid ){
            reset( true );
@@ -56,7 +55,6 @@ namespace mongo {
            nObjects = nDeleted;
        }
        LastError() {
-            overridenById = false;
            reset();
        }
        void reset( bool _valid = false ) {
@@ -127,7 +125,7 @@ namespace mongo {
        
        /** when db receives a message/request, call this */
        void startRequest( Message& m , LastError * connectionOwned );
-        LastError * startRequest( Message& m , int clientId = 0 );
+        LastError * startRequest( Message& m , int clientId );
        
        void disconnect( int clientId );

--- a/db/mr.cpp
+++ b/db/mr.cpp
@@ -176,11 +176,11 @@ namespace mongo {
                { // query options
                    if ( cmdObj["query"].type() == Object ){
                        filter = cmdObj["query"].embeddedObjectUserCheck();
-                        q = filter;
                    }
                    
-                    if ( cmdObj["sort"].type() == Object )
-                        q.sort( cmdObj["sort"].embeddedObjectUserCheck() );
+                    if ( cmdObj["sort"].type() == Object ){
+                        sort = cmdObj["sort"].embeddedObjectUserCheck();
+                    }

                    if ( cmdObj["limit"].isNumber() )
                        limit = cmdObj["limit"].numberLong();
@@ -222,7 +222,7 @@ namespace mongo {
            // query options
            
            BSONObj filter;
-            Query q;
+            BSONObj sort;
            long long limit;

            // functions
@@ -444,7 +444,7 @@ namespace mongo {
                        readlock lock( mr.ns );
                        Client::Context ctx( mr.ns );
                        
-                        shared_ptr<Cursor> temp = bestGuessCursor( mr.ns.c_str(), mr.filter, BSONObj() );
+                        shared_ptr<Cursor> temp = bestGuessCursor( mr.ns.c_str(), mr.filter, mr.sort );
                        auto_ptr<ClientCursor> cursor( new ClientCursor( QueryOption_NoCursorTimeout , temp , mr.ns.c_str() ) );

                        Timer mt;
--- a/db/namespace.cpp
+++ b/db/namespace.cpp
@@ -637,6 +637,8 @@ namespace mongo {
 	}

    bool legalClientSystemNS( const string& ns , bool write ){
+        if( ns == "local.system.replset" ) return true;
+
        if ( ns.find( ".system.users" ) != string::npos )
            return true;

--- a/db/pdfile.cpp
+++ b/db/pdfile.cpp
@@ -728,8 +728,11 @@ namespace mongo {
            try { 
                assert( dropIndexes(d, name.c_str(), "*", errmsg, result, true) );
            }
-            catch( DBException& ) {
-                uasserted(12503,"drop: dropIndexes for collection failed - consider trying repair");
+            catch( DBException& e ) {
+                stringstream ss;
+                ss << "drop: dropIndexes for collection failed - consider trying repair ";
+                ss << " cause: " << e.what();
+                uasserted(12503,ss.str());
            }
            assert( d->nIndexes == 0 );
        }
@@ -1892,7 +1895,10 @@ namespace mongo {
        bb.done();
        if( nNotClosed )
            result.append("nNotClosed", nNotClosed);
-        
+        else {
+            ClientCursor::assertNoCursors();
+        }
+
        return true;
    }
    
--- a/db/query.cpp
+++ b/db/query.cpp
@@ -654,11 +654,12 @@ namespace mongo {
                if ( !ClientCursor::recoverFromYield( _yieldData ) ) {
                    _c.reset();
                    _cc.reset();
+                    _so.reset();
                    massert( 13338, "cursor dropped during query", false );
                    // TODO maybe we want to prevent recording the winning plan as well?
                } 
            }
-        }        
+        }
        
        virtual void next() {
            if ( _findingStartCursor.get() ) {
@@ -772,7 +773,8 @@ namespace mongo {
                _n = _inMemSort ? _so->size() : _n;
            } 
            else if ( _inMemSort ) {
-                _so->fill( _buf, _pq.getFields() , _n );
+                if( _so.get() )
+                    _so->fill( _buf, _pq.getFields() , _n );
            }
            
            if ( _pq.hasOption( QueryOption_CursorTailable ) && _pq.getNumToReturn() != 1 )
--- a/db/repl.cpp
+++ b/db/repl.cpp
@@ -290,11 +290,13 @@ namespace mongo {
            }
        }
        else {
-            result.append("ismaster", replSettings.master || replSettings.slave == 0 ? 1 : 0);
-            //result.append("msg", "not paired");
+            result.appendBool("ismaster", _isMaster() );
        }
-        
-        if ( level ){
+
+        if ( level && replSet ){
+            result.append( "info" , "is replica set" );
+        }
+        else if ( level ){
            BSONObjBuilder sources( result.subarrayStart( "sources" ) );
            
            readlock lk( "local.sources" );
@@ -856,11 +858,13 @@ namespace mongo {
       see logOp() comments.
    */
    void ReplSource::sync_pullOpLog_applyOperation(BSONObj& op, OpTime *localLogTail) {
-        log( 6 ) << "processing op: " << op << endl;
+        if( logLevel >= 6 ) // op.tostring is expensive so doing this check explicitly
+            log(6) << "processing op: " << op << endl;
+
        // skip no-op
-        if ( op.getStringField( "op" )[ 0 ] == 'n' )
+        if( op.getStringField("op")[0] == 'n' )
            return;
-        
+
        char clientName[MaxDatabaseLen];
        const char *ns = op.getStringField("ns");
        nsToDatabase(ns, clientName);
@@ -1301,12 +1305,15 @@ namespace mongo {
                   1) find most recent op in local log
                   2) more()?
                */
-                if ( !oplogReader.more() ) {
+
+                bool moreInitialSyncsPending = !addDbNextPass.empty() && n; // we need "&& n" to assure we actually process at least one op to get a sync point recorded in the first place.
+
+                if ( moreInitialSyncsPending || !oplogReader.more() ) {
                    dblock lk;
                    OpTime nextLastSaved = nextLastSavedLocalTs();
                    {
                        dbtemprelease t;
-                        if ( oplogReader.more() ) {
+                        if ( !moreInitialSyncsPending && oplogReader.more() ) {
                            if ( getInitialSyncCompleted() ) { // if initial sync hasn't completed, break out of loop so we can set to completed or clone more dbs
                                continue;
                            }
@@ -1323,6 +1330,8 @@ namespace mongo {
                    log() << "repl:  end sync_pullOpLog syncedTo: " << syncedTo.toStringLong() << endl;
                    break;
                }
+                else {
+                }

                OCCASIONALLY if( n > 0 && ( n > 100000 || time(0) - saveLast > 60 ) ) { 
 					// periodically note our progress, in case we are doing a lot of work and crash
--- a/db/repl/health.cpp
+++ b/db/repl/health.cpp
@@ -171,7 +171,7 @@ namespace mongo {
    }

    void ReplSetImpl::_getOplogDiagsAsHtml(unsigned server_id, stringstream& ss) const { 
-        Member *m = findById(server_id);
+        const Member *m = findById(server_id);
        if( m == 0 ) { 
            ss << "Error : can't find a member with id: " << server_id << '\n';
            return;
@@ -329,7 +329,7 @@ namespace mongo {
        _rsLog.toHTML( s );
    }

-    Member* ReplSetImpl::findById(unsigned id) const { 
+    const Member* ReplSetImpl::findById(unsigned id) const { 
        if( id == _self->id() ) return _self;
        for( Member *m = head(); m; m = m->next() )
            if( m->id() == id ) 
@@ -371,6 +371,7 @@ namespace mongo {
            v.push_back(bb.obj());
            m = m->next();
        }
+        sort(v.begin(), v.end());
        b.append("set", name());
        b.appendTimeT("date", time(0));
        b.append("myState", box.getState().s);
--- a/db/repl/replset_commands.cpp
+++ b/db/repl/replset_commands.cpp
@@ -26,7 +26,7 @@

 namespace mongo { 

-    void checkAllMembersUpForConfigChange(const ReplSetConfig& cfg, bool initial);
+    void checkMembersUpForConfigChange(const ReplSetConfig& cfg, bool initial);

    /* commands in other files:
         replSetHeartbeat - health.cpp
@@ -151,9 +151,9 @@ namespace mongo {
                    return false;
                }

-                checkAllMembersUpForConfigChange(newConfig,false);
+                checkMembersUpForConfigChange(newConfig,false);

-                log() << "replSet replSetReconfig all members seem up" << rsLog;
+                log() << "replSet replSetReconfig [2]" << rsLog;

                theReplSet->haveNewConfig(newConfig, true);
                ReplSet::startupStatusMsg = "replSetReconfig'd";
--- a/db/repl/rs.cpp
+++ b/db/repl/rs.cpp
@@ -123,6 +123,7 @@ namespace mongo {
    void ReplSetImpl::_fillIsMaster(BSONObjBuilder& b) {
        const StateBox::SP sp = box.get();
        bool isp = sp.state.primary();
+        b.append("setName", name());
        b.append("ismaster", isp);
        b.append("secondary", sp.state.secondary());
        {
@@ -390,9 +391,9 @@ namespace mongo {
                        startupStatus = EMPTYCONFIG;
                        startupStatusMsg = "can't get " + rsConfigNs + " config from self or any seed (EMPTYCONFIG)";
                        log() << "replSet can't get " << rsConfigNs << " config from self or any seed (EMPTYCONFIG)" << rsLog;
-                        log() << "replSet   have you ran replSetInitiate yet?" << rsLog;
+                        log(1) << "replSet have you ran replSetInitiate yet?" << rsLog;
                        if( _seeds->size() == 0 )
-                            log() << "replSet   no seed hosts were specified on the --replSet command line - that might be the issue" << rsLog;
+                            log(1) << "replSet info no seed hosts were specified on the --replSet command line" << rsLog;
                    }
                    else {
                        startupStatus = EMPTYUNREACHABLE;
--- a/db/repl/rs.h
+++ b/db/repl/rs.h
@@ -305,7 +305,9 @@ namespace mongo {

    private:
        Member* head() const { return _members.head(); }
-        Member* findById(unsigned id) const;
+    public:
+        const Member* findById(unsigned id) const;
+    private:
        void _getTargets(list<Target>&, int &configVersion);
        void getTargets(list<Target>&, int &configVersion);
        void startThreads();
--- a/db/repl/rs_config.cpp
+++ b/db/repl/rs_config.cpp
@@ -114,7 +114,7 @@ namespace mongo {
        mchk(_id >= 0 && _id <= 255);
        mchk(priority >= 0 && priority <= 1000);
        mchk(votes >= 0 && votes <= 100);
-        uassert(13419, "replica set config : this version of mongo only supports priorities 0 and 1", priority == 0 || priority == 1);
+        uassert(13419, "this version of mongod only supports priorities 0 and 1", priority == 0 || priority == 1);
    }

    /*static*/ bool ReplSetConfig::legalChange(const ReplSetConfig& o, const ReplSetConfig& n, string& errmsg) { 
@@ -133,7 +133,7 @@ namespace mongo {

        /* TODO : MORE CHECKS HERE */

-        cout << "TODO : don't allow removal of a node until we handle it at the removed node end." << endl;
+        log() << "replSet TODO : don't allow removal of a node until we handle it at the removed node end?" << endl;
        // we could change its votes to zero perhaps instead as a short term...

        return true;
@@ -220,7 +220,7 @@ namespace mongo {
            catch(DBException& e) { 
                log() << "replSet cfg parsing exception for members[" << i << "] " << e.what() << rsLog;
                stringstream ss;
-                ss << "replSet members[" << i << "] bad config object";
+                ss << "bad config for member[" << i << "] " << e.what();
                uassert(13135, ss.str(), false);
            }
            if( !(ords.count(m._id) == 0 && hosts.count(m.h.toString()) == 0) ) {
--- a/db/repl/rs_initiate.cpp
+++ b/db/repl/rs_initiate.cpp
@@ -32,10 +32,12 @@ using namespace mongoutils;

 namespace mongo { 

-    /* throws 
-       @param initial - 
+    /* called on a reconfig AND on initiate
+       throws 
+       @param initial true when initiating
    */
-    void checkAllMembersUpForConfigChange(const ReplSetConfig& cfg, bool initial) {
+    void checkMembersUpForConfigChange(const ReplSetConfig& cfg, bool initial) {
+        int failures = 0;
        int me = 0;
        for( vector<ReplSetConfig::MemberCfg>::const_iterator i = cfg.members.begin(); i != cfg.members.end(); i++ ) {
            if( i->h.isSelf() ) {
@@ -45,7 +47,7 @@ namespace mongo {
                }
            }
        }
-        uassert(13278, "bad config?", me <= 1);
+        uassert(13278, "bad config - dups?", me <= 1); // dups?
        uassert(13279, "can't find self in the replset config", me == 1);

        for( vector<ReplSetConfig::MemberCfg>::const_iterator i = cfg.members.begin(); i != cfg.members.end(); i++ ) {
@@ -62,12 +64,11 @@ namespace mongo {
                    }
                }
                catch(DBException& e) { 
-                    log() << "replSet requestHeartbeat " << i->h.toString() << " : " << e.toString() << rsLog;
+                    log() << "replSet cmufcc requestHeartbeat " << i->h.toString() << " : " << e.toString() << rsLog;
                }
                catch(...) { 
-                    log() << "replSet error exception in requestHeartbeat?" << rsLog;
+                    log() << "replSet cmufcc error exception in requestHeartbeat?" << rsLog;
                }
-                cout << "TEMP hb res cfg change:" << res.toString() << endl;
                if( res.getBoolField("mismatch") )
                    uasserted(13145, "set name does not match the set name host " + i->h.toString() + " expects");
                if( *res.getStringField("set") ) {
@@ -83,9 +84,31 @@ namespace mongo {
                    }
                }
                if( !ok && !res["rs"].trueValue() ) {
-                    if( !res.isEmpty() )
+                    if( !res.isEmpty() ) {
+                        /* strange.  got a response, but not "ok". log it. */
                        log() << "replSet warning " << i->h.toString() << " replied: " << res.toString() << rsLog;
-                    uasserted(13144, "need members up to initiate/reconfig, not ok: " + i->h.toString());
+                    }
+
+                    bool allowFailure = false;
+                    failures++;
+                    if( res.isEmpty() && !initial && failures == 1 ) {
+                        /* for now we are only allowing 1 node to be down on a reconfig.  this can be made to be a minority
+                           trying to keep change small as release is near.
+                           */
+                        const Member* m = theReplSet->findById( i->_id );
+                        if( m ) { 
+                            // ok, so this was an existing member (wouldn't make sense to add to config a new member that is down)
+                            assert( m->h().toString() == i->h.toString() );
+                            allowFailure = true;
+                        }
+                    }
+
+                    if( !allowFailure ) {
+                        string msg = string("need members up to initiate, not ok : ") + i->h.toString();
+                        if( !initial )
+                            msg = string("need most members up to reconfigure, not ok : ") + i->h.toString();
+                        uasserted(13144, msg);
+                    }
                }
            }
            if( initial ) {
@@ -175,8 +198,10 @@ namespace mongo {
                configObj = cmdObj["replSetInitiate"].Obj();
            }

+            bool parsed = false;
            try {
                ReplSetConfig newConfig(configObj);
+                parsed = true;

                if( newConfig.version > 1 ) { 
                    errmsg = "can't initiate with a version number greater than 1";
@@ -185,7 +210,7 @@ namespace mongo {

                log() << "replSet replSetInitiate config object parses ok, " << newConfig.members.size() << " members specified" << rsLog;

-                checkAllMembersUpForConfigChange(newConfig, true);
+                checkMembersUpForConfigChange(newConfig, true);

                log() << "replSet replSetInitiate all members seem up" << rsLog;

@@ -199,7 +224,11 @@ namespace mongo {
            }
            catch( DBException& e ) { 
                log() << "replSet replSetInitiate exception: " << e.what() << rsLog;
-                throw;
+                if( !parsed ) 
+                    errmsg = string("couldn't parse cfg object ") + e.what();
+                else
+                    errmsg = string("couldn't initiate : ") + e.what();
+                return false;
            }

            return true;
--- a/db/repl/rs_sync.cpp
+++ b/db/repl/rs_sync.cpp
@@ -203,8 +203,8 @@ namespace mongo {
            OpTime ts = o["ts"]._opTime();
            long long h = o["h"].numberLong();
            if( ts != lastOpTimeWritten || h != lastH ) { 
-                log() << "TEMP our last op time written: " << lastOpTimeWritten.toStringPretty() << endl;
-                log() << "TEMP primary's GTE: " << ts.toStringPretty() << endl;
+                log(1) << "TEMP our last op time written: " << lastOpTimeWritten.toStringPretty() << endl;
+                log(1) << "TEMP primary's GTE: " << ts.toStringPretty() << endl;
                /*
                }*/

@@ -243,7 +243,6 @@ namespace mongo {
                        }
                        else { 
                            sethbmsg(str::stream() << "still syncing, not yet to minValid optime " << minvalid.toString());
-                            //log() << "TEMP " << lastOpTimeWritten.toString() << rsLog;
                        }

                        /* todo: too stale capability */
--- a/db/replpair.h
+++ b/db/replpair.h
@@ -164,7 +164,7 @@ namespace mongo {
        if( replSet ) {
            /* todo: speed up the secondary case.  as written here there are 2 mutex entries, it can be 1. */
            if( isMaster() ) return;
-            notMasterUnless( pq.hasOption(QueryOption_SlaveOk) && theReplSet->isSecondary() );
+            notMasterUnless( pq.hasOption(QueryOption_SlaveOk) && theReplSet && theReplSet->isSecondary() );
        } else {
            notMasterUnless(isMaster() || pq.hasOption(QueryOption_SlaveOk) || replSettings.slave == SimpleSlave );
        }
--- a/db/restapi.cpp
+++ b/db/restapi.cpp
@@ -266,8 +266,8 @@ namespace mongo {
            
            ss << "# databases: " << dbHolder.size() << '\n';
            
-            if( ClientCursor::byLocSize()>500 )
-                ss << "Cursors byLoc.size(): " << ClientCursor::byLocSize() << '\n';
+            if( ClientCursor::numCursors()>500 )
+                ss << "# Cursors: " << ClientCursor::numCursors() << '\n';
            
            ss << "\nreplication: ";
            if( *replInfo )
--- a/db/scanandorder.h
+++ b/db/scanandorder.h
@@ -88,16 +88,17 @@ namespace mongo {

        void _add(BSONObj& k, BSONObj o, DiskLoc* loc) {
            if (!loc){
-                best.insert(make_pair(k,o));
+                best.insert(make_pair(k.getOwned(),o.getOwned()));
            } else {
                BSONObjBuilder b;
                b.appendElements(o);
                b.append("$diskLoc", loc->toBSONObj());
-                best.insert(make_pair(k, b.obj()));
+                best.insert(make_pair(k.getOwned(), b.obj().getOwned()));
            }
        }

        void _addIfBetter(BSONObj& k, BSONObj o, BestMap::iterator i, DiskLoc* loc) {
+            /* todo : we don't correct approxSize here. */
            const BSONObj& worstBestKey = i->first;
            int c = worstBestKey.woCompare(k, order.pattern);
            if ( c > 0 ) {
@@ -124,7 +125,11 @@ namespace mongo {
            BSONObj k = order.getKeyFromObject(o);
            if ( (int) best.size() < limit ) {
                approxSize += k.objsize();
-                uassert( 10128 ,  "too much key data for sort() with no index.  add an index or specify a smaller limit", approxSize < 1 * 1024 * 1024 );
+                approxSize += o.objsize();
+
+                /* note : adjust when bson return limit adjusts. note this limit should be a bit higher. */
+                uassert( 10128 ,  "too much data for sort() with no index.  add an index or specify a smaller limit", approxSize < 32 * 1024 * 1024 );
+
                _add(k, o, loc);
                return;
            }
--- a/db/update.cpp
+++ b/db/update.cpp
@@ -527,7 +527,7 @@ namespace mongo {
            string field = root + e.fieldName();
            FieldCompareResult cmp = compareDottedFieldNames( m->second.m->fieldName , field );

-            DEBUGUPDATE( "\t\t\t field:" << field << "\t mod:" << m->second.m->fieldName << "\t cmp:" << cmp );
+            DEBUGUPDATE( "\t\t\t field:" << field << "\t mod:" << m->second.m->fieldName << "\t cmp:" << cmp << "\t short: " << e.fieldName() );
            
            switch ( cmp ){
                
@@ -550,6 +550,13 @@ namespace mongo {
                    e = es.next();
                    m++;
                }
+                else {
+                    // this is a very weird case
+                    // have seen it in production, but can't reproduce
+                    // this assert prevents an inf. loop
+                    // but likely isn't the correct solution
+                    assert(0);
+                }
                continue;
            }
            case LEFT_BEFORE: // Mod on a field that doesn't exist
@@ -613,6 +620,8 @@ namespace mongo {
            BSONObjIteratorSorted i( query );
            while ( i.more() ){
                BSONElement e = i.next();
+                if ( e.fieldName()[0] == '$' ) // for $atomic and anything else we add
+                    continue;

                if ( e.type() == Object && e.embeddedObject().firstElement().fieldName()[0] == '$' ){
                    // this means this is a $gt type filter, so don't make part of the new object
--- a/dbtests/jsobjtests.cpp
+++ b/dbtests/jsobjtests.cpp
@@ -389,6 +389,35 @@ namespace JsobjTests {
            }
        };

+        class ToStringNumber {
+        public:
+            
+            void run(){
+                BSONObjBuilder b;
+                b.append( "a" , (int)4 );
+                b.append( "b" , (double)5 );
+                b.append( "c" , (long long)6 );
+                
+                b.append( "d" , 123.456789123456789123456789123456789 );
+                b.append( "e" , 123456789.123456789123456789123456789 );
+                b.append( "f" , 1234567891234567891234.56789123456789 );
+
+                b.append( "g" , -123.456 );
+                
+                BSONObj x = b.obj();
+                ASSERT_EQUALS( "4", x["a"].toString( false , true ) );
+                ASSERT_EQUALS( "5.0", x["b"].toString( false , true ) );
+                ASSERT_EQUALS( "6", x["c"].toString( false , true ) );                
+
+                ASSERT_EQUALS( "123.4567891234568" , x["d"].toString( false , true ) );
+                ASSERT_EQUALS( "123456789.1234568" , x["e"].toString( false , true ) );
+                // ASSERT_EQUALS( "1.234567891234568e+21" , x["f"].toString( false , true ) ); // windows and *nix are different - TODO, work around for test or not bother?
+                
+                ASSERT_EQUALS( "-123.456" , x["g"].toString( false , true ) );
+
+            }
+        };
+
        class NullString {
        public:
            void run() {
@@ -1693,6 +1722,7 @@ namespace JsobjTests {
            add< BSONObjTests::AppendIntOrLL >();
            add< BSONObjTests::AppendNumber >();
            add< BSONObjTests::ToStringArray >();
+            add< BSONObjTests::ToStringNumber >();
            add< BSONObjTests::NullString >();
            add< BSONObjTests::Validation::BadType >();
            add< BSONObjTests::Validation::EooBeforeEnd >();
--- a/debian/changelog
+++ b/debian/changelog
@@ -1,8 +1,33 @@
+mongodb (1.6.2) unstable; urgency=low
+
+  * replica_sets some fixes
+  * sharding some fixes with rs
+  * full change log http://jira.mongodb.org/browse/SERVER/fixforversion/10187
+
+  -- Richard Kreuter <richard@10gen.com>  Wed, 1 Sep 2010 16:56:28 -0500
+
+
+mongodb (1.6.1) unstable; urgency=low
+
+  * replica_sets some fixes
+  * sharding some fixes with rs
+  * full change log http://jira.mongodb.org/browse/SERVER/fixforversion/10183
+
+  -- Richard Kreuter <richard@10gen.com>  Tue, 17 Aug 2010 16:56:28 -0500
+
+mongodb (1.6.0) unstable; urgency=low
+
+  * sharding stable
+  * replica_sets stable
+
+  -- Richard Kreuter <richard@10gen.com>  Thu, 05 Aug 2010 16:56:28 -0500
+
 mongodb (1.5.8) unstable; urgency=low

  * sharding lots of changes
  * replica_sets lots of changes

+  -- Richard Kreuter <richard@10gen.com>  Tue, 03 Aug 2010 16:56:28 -0500

 mongodb (1.5.7) unstable; urgency=low

--- a/2
+++ b/2
@@ -3,7 +3,7 @@
 #---------------------------------------------------------------------------
 DOXYFILE_ENCODING      = UTF-8
 PROJECT_NAME           = MongoDB
-PROJECT_NUMBER         = 1.5.8
+PROJECT_NUMBER         = 1.6.2
 OUTPUT_DIRECTORY       = docs/doxygen
 CREATE_SUBDIRS         = NO
 OUTPUT_LANGUAGE        = English
--- a/jstests/basic3.js
+++ b/jstests/basic3.js
@@ -1,5 +1,5 @@

-t = db.getCollection( "foo" );
+t = db.getCollection( "foo_basic3" );

 t.find( { "a.b" : 1 } ).toArray();

--- a/jstests/basic9.js
+++ b/jstests/basic9.js
@@ -1,5 +1,5 @@

-t = db.getCollection( "foo" );
+t = db.getCollection( "foo_basic9" );

 t.save( { "foo$bar" : 5 } );

--- a/jstests/dbadmin.js
+++ b/jstests/dbadmin.js
@@ -21,4 +21,10 @@ assert( res.databases && res.databases.length > 0 , "listDatabases 1 " + tojson(
 x = db._adminCommand( "ismaster" );
 assert( x.ismaster , "ismaster failed: " + tojson( x ) )

+before = db.runCommand( "serverStatus" )
+sleep( 5000 )
+after = db.runCommand( "serverStatus" )
+assert.lt( 3 , after.uptimeEstimate , "up1" )
+assert.gt( after.uptimeEstimate , before.uptimeEstimate , "up2" )
+
 // TODO: add more tests here
--- a/jstests/dbcase.js
+++ b/jstests/dbcase.js
@@ -1,6 +1,6 @@

-a = db.getSisterDB( "test_dbnamea" )
-b = db.getSisterDB( "test_dbnameA" )
+a = db.getSisterDB( "dbcasetest_dbnamea" )
+b = db.getSisterDB( "dbcasetest_dbnameA" )

 a.dropDatabase();
 b.dropDatabase();
--- a/jstests/delx.js
+++ b/jstests/delx.js
@@ -0,0 +1,31 @@
+
+a = db.getSisterDB("delxa" )
+b = db.getSisterDB("delxb" )
+
+function setup( mydb ){
+    mydb.dropDatabase();
+    for ( i=0; i<100; i++ ){
+        mydb.foo.insert( { _id : i } );
+    }
+    mydb.getLastError();
+}
+
+setup( a );
+setup( b );
+
+assert.eq( 100 , a.foo.find().itcount() , "A1" )
+assert.eq( 100 , b.foo.find().itcount() , "A2" )
+
+x = a.foo.find().sort( { _id : 1 } ).batchSize( 60 )
+y = b.foo.find().sort( { _id : 1 } ).batchSize( 60 )
+
+x.next();
+y.next();
+
+a.foo.remove( { _id : { $gt : 50 } } );
+
+assert.eq( 51 , a.foo.find().itcount() , "B1" )
+assert.eq( 100 , b.foo.find().itcount() , "B2" )
+
+assert.eq( 59 , x.itcount() , "C1" ) 
+assert.eq( 99 , y.itcount() , "C2" );  // this was asserting because ClientCursor byLoc doesn't take db into consideration
--- a/jstests/repl/basic1.js
+++ b/jstests/repl/basic1.js
@@ -141,6 +141,14 @@ assert.soon( function(){ return am.lotOfIndexes.getIndexes().length == as.lotOfI

 assert.eq( am.lotOfIndexes.getIndexes().length , as.lotOfIndexes.getIndexes().length , "lots of indexes b" )

+// multi-update with $inc
+
+am.mu1.update( { _id : 1 , $atomic : 1 } , { $inc : { x : 1 } } , true , true )
+x = { _id : 1 , x : 1 }
+assert.eq( x , am.mu1.findOne() , "mu1" );
+assert.soon( function(){ z = as.mu1.findOne(); printjson( z ); return friendlyEqual( x , z ); } , "mu2" )
+
+

 rt.stop();

--- a/jstests/replsets/replsetrestart2.js
+++ b/jstests/replsets/replsetrestart2.js
@@ -27,7 +27,7 @@ doTest = function( signal ) {
    var master = replTest.getMaster();
    var config = master.getDB("local").system.replset.findOne();

-    replTest.stopSet( signal );
+    replTest.stopSet( signal , true );

    replTest.restart(0);
    replTest.restart(1);
--- a/jstests/sharding/addshard2.js
+++ b/jstests/sharding/addshard2.js
@@ -0,0 +1,52 @@
+
+// don't start any shards, yet
+s = new ShardingTest( "add_shard2", 1, 0, 1, {useHostname : true} );
+
+var conn1 = startMongodTest( 30001 , "add_shard21" , 0 , {useHostname : true} );
+var conn2 = startMongodTest( 30002 , "add_shard22" , 0 , {useHostname : true} );
+
+var rs1 = new ReplSetTest( { "name" : "add_shard2_rs1", nodes : 3 , startPort : 31200 } );
+rs1.startSet();
+rs1.initiate();
+var master1 = rs1.getMaster();
+
+var rs2 = new ReplSetTest( { "name" : "add_shard2_rs2", nodes : 3 , startPort : 31203 } );
+rs2.startSet();
+rs2.initiate();
+var master2 = rs2.getMaster();
+
+// step 1. name given
+assert(s.admin.runCommand({"addshard" : getHostName()+":30001", "name" : "bar"}).ok, "failed to add shard in step 1");
+var shard = s.getDB("config").shards.findOne({"_id" : {"$nin" : ["shard0000"]}});
+assert(shard, "shard wasn't found");
+assert.eq("bar", shard._id, "shard has incorrect name");
+
+// step 2. replica set
+assert(s.admin.runCommand({"addshard" : "add_shard2_rs1/"+getHostName()+":31200"}).ok, "failed to add shard in step 2");
+shard = s.getDB("config").shards.findOne({"_id" : {"$nin" : ["shard0000", "bar"]}});
+assert(shard, "shard wasn't found");
+assert.eq("add_shard2_rs1", shard._id, "t2 name");
+
+// step 3. replica set w/ name given
+assert(s.admin.runCommand({"addshard" : "add_shard2_rs2/"+getHostName()+":31203", "name" : "myshard"}).ok,
+       "failed to add shard in step 4");
+shard = s.getDB("config").shards.findOne({"_id" : {"$nin" : ["shard0000", "bar", "add_shard2_rs1"]}});
+assert(shard, "shard wasn't found");
+assert.eq("myshard", shard._id, "t3 name");
+
+// step 4. no name given
+assert(s.admin.runCommand({"addshard" : getHostName()+":30002"}).ok, "failed to add shard in step 4");
+shard = s.getDB("config").shards.findOne({"_id" : {"$nin" : ["shard0000", "bar", "add_shard2_rs1", "myshard"]}});
+assert(shard, "shard wasn't found");
+assert.eq("shard0001", shard._id, "t4 name");
+
+assert.eq(s.getDB("config").shards.count(), 5, "unexpected number of shards");
+
+// step 5. replica set w/ a wrong host
+assert(!s.admin.runCommand({"addshard" : "add_shard2_rs2/NonExistingHost:31203"}).ok, "accepted bad hostname in step 5");
+
+// step 6. replica set w/ mixed wrong/right hosts
+assert(!s.admin.runCommand({"addshard" : "add_shard2_rs2/"+getHostName()+":31203,foo:9999"}).ok,
+       "accepted bad hostname in step 6");
+
+s.stop();
--- a/jstests/sharding/sort1.js
+++ b/jstests/sharding/sort1.js
@@ -6,8 +6,14 @@ s.adminCommand( { shardcollection : "test.data" , key : { num : 1 } } );

 db = s.getDB( "test" );

-for ( i=0; i<100; i++ ){
-    db.data.insert( { _id : i , num : i } );
+N = 100
+
+forward = []
+backward = [] 
+for ( i=0; i<N; i++ ){
+    db.data.insert( { _id : i , num : i , x : N - i } )
+    forward.push( i )
+    backward.push( ( N - 1 ) - i )
 }
 db.getLastError();

@@ -23,7 +29,7 @@ assert.eq( temp[0].shard , temp[2].shard , "A2" );
 assert.neq( temp[0].shard , temp[1].shard , "A3" );

 temp = db.data.find().sort( { num : 1 } ).toArray();
-assert.eq( 100 , temp.length , "B1" );
+assert.eq( N , temp.length , "B1" );
 for ( i=0; i<100; i++ ){
    assert.eq( i , temp[i].num , "B2" )
 }
@@ -39,6 +45,37 @@ assert.eq( 67 , z.length , "C2" )

 print( "a: " + a + " b:" + b + " mongos slow down: " + Math.ceil( 100 * ( ( a - b ) / b ) ) + "%" )

-//assert.lt( a , b * 1.3 , "C3 - speed test" );
+// -- secondary index sorting
+
+function getSorted( by , want , dir , proj ){
+    var s = {}
+    s[by] = dir || 1;
+    printjson( s )
+    var cur = db.data.find( {} , proj || {} ).sort( s )
+    return terse( cur.map( function(z){ return z[want]; } ) );
+}
+
+function terse( a ){
+    var s = "";
+    for ( var i=0; i<a.length; i++ ){
+        if ( i > 0 )
+            s += ",";
+        s += a[i];
+    }
+    return s;
+}
+
+forward = terse(forward);
+backward = terse(backward);
+
+assert.eq( forward , getSorted( "num" , "num" , 1 ) , "D1" )
+assert.eq( backward , getSorted( "num" , "num" , -1 ) , "D2" )
+
+assert.eq( backward , getSorted( "x" , "num" , 1 ) , "D3" )
+assert.eq( forward , getSorted( "x" , "num" , -1 ) , "D4" )
+
+assert.eq( backward , getSorted( "x" , "num" , 1 , { num : 1 } ) , "D5" )
+assert.eq( forward , getSorted( "x" , "num" , -1 , { num : 1 } ) , "D6" )
+

 s.stop();
--- a/jstests/slowNightly/capped4.js
+++ b/jstests/slowNightly/capped4.js
@@ -25,4 +25,10 @@ assert( t.findOne( { i : i - 1 } ), "E" );
 t.remove( { i : i - 1 } );
 assert( db.getLastError().indexOf( "capped" ) >= 0, "F" );

-assert( t.validate().valid, "G" );
+assert( t.validate().valid, "G" );
+
+/* there is a cursor open here, so this is a convenient place for a quick cursor test. */
+
+db._adminCommand("closeAllDatabases");
+
+//assert( db.serverStatus().cursors.totalOpen == 0, "cursors open and shouldn't be");
--- a/jstests/slowNightly/cursor8.js
+++ b/jstests/slowNightly/cursor8.js
--- a/jstests/slowNightly/recstore.js
+++ b/jstests/slowNightly/recstore.js
--- a/jstests/slowNightly/sharding_rs1.js
+++ b/jstests/slowNightly/sharding_rs1.js
@@ -15,7 +15,7 @@ while ( bigString.length < 10000 )
 inserted = 0;
 num = 0;
 while ( inserted < ( 20 * 1024 * 1024 ) ){
-    db.foo.insert( { _id : num++ , s : bigString } );
+    db.foo.insert( { _id : num++ , s : bigString , x : Math.random() } );
    inserted += bigString.length;
 }

@@ -48,6 +48,7 @@ assert.soon( function(){
    return d < 5;
 } , "balance didn't happen" , 1000 * 60 * 3 , 5000 );

+s.config.settings.update( { _id: "balancer" }, { $set : { stopped: true } } , true );

 for ( i=0; i<s._rs.length; i++ ){
    r = s._rs[i];
@@ -58,4 +59,14 @@ for ( i=0; i<s._rs.length; i++ ){
        assert.eq( x.master.md5 , x.slaves[j].md5 , "hashes same for: " + r.url + " slave: " + j );
 }

+
+assert.eq( num , db.foo.find().count() , "C1" )
+assert.eq( num , db.foo.find().itcount() , "C2" )
+assert.eq( num , db.foo.find().sort( { _id : 1 } ).itcount() , "C3" )
+assert.eq( num , db.foo.find().sort( { _id : -1 } ).itcount() , "C4" )
+db.foo.ensureIndex( { x : 1 } )
+assert.eq( num , db.foo.find().sort( { x : 1 } ).itcount() , "C5" )
+assert.eq( num , db.foo.find().sort( { x : -1 } ).itcount() , "C6" )
+
+
 s.stop()
--- a/rpm/mongo.spec
+++ b/rpm/mongo.spec
@@ -1,5 +1,5 @@
 Name: mongo
-Version: 1.5.8
+Version: 1.6.2
 Release: mongodb_1%{?dist}
 Summary: mongo client shell and tools
 License: AGPL 3.0
@@ -130,11 +130,6 @@ fi
 %attr(0755,mongod,mongod) %dir /var/log/mongo
 %attr(0640,mongod,mongod) %config(noreplace) %verify(not md5 size mtime) /var/log/mongo/mongod.log

-%files devel
-/usr/include/mongo
-%{_libdir}/libmongoclient.a
-#%{_libdir}/libmongotestfiles.a
-
 %changelog
 * Thu Jan 28 2010 Richard M Kreuter <richard@10gen.com>
 - Minor fixes.
--- a/s/chunk.cpp
+++ b/s/chunk.cpp
@@ -265,7 +265,7 @@ namespace mongo {
        }

        // Save the new key boundaries in the configDB.
-        _manager->save();
+        _manager->save( false );

        // Log all these changes in the configDB's log. We log a simple split differently than a multi-split.
        if ( newChunks.size() == 1) {
@@ -584,7 +584,7 @@ namespace mongo {

            _shards.insert(c->getShard());

-            save_inlock();
+            save_inlock( true );
            log() << "no chunks for:" << ns << " so creating first: " << c->toString() << endl;
        }
    }
@@ -880,16 +880,18 @@ namespace mongo {
        configServer.logChange( "dropCollection" , _ns , BSONObj() );
    }
    
-    void ChunkManager::save(){
+    void ChunkManager::save( bool major ){
        rwlock lk( _lock , true ); 
-        save_inlock();
+        save_inlock( major );
    }
    
-    void ChunkManager::save_inlock(){
+    void ChunkManager::save_inlock( bool major ){
        
        ShardChunkVersion a = getVersion_inlock();
        assert( a > 0 || _chunkMap.size() <= 1 );
-        ShardChunkVersion nextChunkVersion = a.incMajor();
+        ShardChunkVersion nextChunkVersion = a;
+        nextChunkVersion.inc( major );
+
        vector<ChunkPtr> toFix;
        vector<ShardChunkVersion> newVersions;
        
@@ -907,7 +909,7 @@ namespace mongo {
            _sequenceNumber = ++NextSequenceNumber;

            ShardChunkVersion myVersion = nextChunkVersion;
-            ++nextChunkVersion;
+            nextChunkVersion.incMinor();
            toFix.push_back( c );
            newVersions.push_back( myVersion );

--- a/s/chunk.h
+++ b/s/chunk.h
@@ -269,7 +269,7 @@ namespace mongo {
        void getAllShards( set<Shard>& all );
        void getShardsForRange(set<Shard>& shards, const BSONObj& min, const BSONObj& max); // [min, max)

-        void save();
+        void save( bool major );

        string toString() const;

@@ -307,7 +307,7 @@ namespace mongo {
        void _reload_inlock();
        void _load();

-        void save_inlock();
+        void save_inlock( bool major );
        ShardChunkVersion getVersion_inlock() const;
        void ensureIndex_inlock();
        
--- a/s/commands_admin.cpp
+++ b/s/commands_admin.cpp
@@ -616,30 +616,44 @@ namespace mongo {
                help << "add a new shard to the system";
            }
            bool run(const string& , BSONObj& cmdObj, string& errmsg, BSONObjBuilder& result, bool){
-                HostAndPort shardAddr( cmdObj.firstElement().valuestrsafe() );
-                if ( shardAddr.isLocalHost() != grid.allowLocalHost() ){
-                    errmsg = "can't use localhost as a shard since all shards need to communicate. "
-                             "either use all shards and configdbs in localhost or all in actual IPs " ;
-                    log() << "addshard request " << cmdObj << " failed: attempt to mix localhosts and IPs" << endl;
+                errmsg.clear();
+
+                // get replica set component hosts
+                ConnectionString servers = ConnectionString::parse( cmdObj.firstElement().valuestrsafe() , errmsg );
+                if ( ! errmsg.empty() ){
+                    log() << "addshard request " << cmdObj << " failed:" << errmsg << endl;
                    return false;
                }

-                // HostAndPort defaults to a different port; adjust if needed.
-                if ( shardAddr.port() == CmdLine::DefaultDBPort ){
-                    shardAddr.setPort( CmdLine::ShardServerPort );
+                // using localhost in server names implies every other process must use locahost addresses too
+                vector<HostAndPort> serverAddrs = servers.getServers();
+                for ( size_t i = 0 ; i < serverAddrs.size() ; i++ ){ 
+                    if ( serverAddrs[i].isLocalHost() != grid.allowLocalHost() ){
+                        errmsg = "can't use localhost as a shard since all shards need to communicate. "
+                                 "either use all shards and configdbs in localhost or all in actual IPs " ;
+                        log() << "addshard request " << cmdObj << " failed: attempt to mix localhosts and IPs" << endl;
+                        return false;
+                    }
+
+                    // it's fine if mongods of a set all use default port
+                    if ( ! serverAddrs[i].hasPort() ){
+                        serverAddrs[i].setPort( CmdLine::ShardServerPort );
+                    }
                }

+                // name is optional; addShard will provide one if needed
                string name = "";
                if ( cmdObj["name"].type() == String ) {
                    name = cmdObj["name"].valuestrsafe();
                } 

+                // maxSize is the space usage cap in a shard in MBs
                long long maxSize = 0;
                if ( cmdObj[ ShardFields::maxSize.name() ].isNumber() ){
                    maxSize = cmdObj[ ShardFields::maxSize.name() ].numberLong();
                }
                
-                if ( ! grid.addShard( &name , shardAddr.toString() , maxSize , errmsg ) ){
+                if ( ! grid.addShard( &name , servers , maxSize , errmsg ) ){
                    log() << "addshard request " << cmdObj << " failed: " << errmsg << endl;
                    return false;
                }
--- a/s/commands_public.cpp
+++ b/s/commands_public.cpp
@@ -745,6 +745,7 @@ namespace mongo {
                         fn == "reduce" ||
                         fn == "query" ||
                         fn == "sort" ||
+                         fn == "scope" ||
                         fn == "verbose" ){
                        b.append( e );
                    }
--- a/s/config.cpp
+++ b/s/config.cpp
@@ -653,7 +653,13 @@ namespace mongo {
        BSONObj msg = BSON( "_id" << id.str() << "server" << getHostNameCached() << "time" << DATENOW <<
                            "what" << what << "ns" << ns << "details" << detail );
        log() << "config change: " << msg << endl;
-        conn->insert( "config.changelog" , msg );
+
+        try {
+            conn->insert( "config.changelog" , msg );
+        }
+        catch ( std::exception& e ){
+            log() << "not logging config change: " << e.what() << endl;                
+        }
        
        conn.done();
    }
--- a/s/d_migrate.cpp
+++ b/s/d_migrate.cpp
@@ -528,7 +528,8 @@ namespace mongo {
                
                log(0) << "_recvChunkStatus : " << res << endl;
                
-                if ( ! ok ){
+                if ( ! ok || res["state"].String() == "fail" ){
+                    log( LL_ERROR ) << "_recvChunkStatus error : " << res << endl;
                    errmsg = "_recvChunkStatus error";
                    result.append( "cause" ,res );
                    return false;
@@ -544,7 +545,7 @@ namespace mongo {
                // 5.a
                migrateFromStatus._inCriticalSection = true;
                ShardChunkVersion myVersion = maxVersion;
-                ++myVersion;
+                myVersion.incMajor();
                
                {
                    dblock lk;
@@ -587,7 +588,8 @@ namespace mongo {
                    if ( ! x.isEmpty() ){
                        
                        BSONObjBuilder temp2;
-                        ++myVersion;
+                        myVersion.incMinor();
+
                        temp2.appendTimestamp( "lastmod" , myVersion );
                        
                        shardingState.setVersion( ns , myVersion );
@@ -685,10 +687,12 @@ namespace mongo {
            catch ( std::exception& e ){
                state = FAIL;
                errmsg = e.what();
+                log( LL_ERROR ) << "migrate failed: " << e.what() << endl;
            }
            catch ( ... ){
                state = FAIL;
                errmsg = "UNKNOWN ERROR";
+                log( LL_ERROR ) << "migrate failed with unknown exception" << endl;
            }
            active = false;
        }
@@ -739,7 +743,7 @@ namespace mongo {
                auto_ptr<DBClientCursor> cursor = conn->query( ns , Query().minKey( min ).maxKey( max ) , /* QueryOption_Exhaust */ 0 );
                assert( cursor.get() );
                while ( cursor->more() ){
-                    BSONObj o = cursor->next();
+                    BSONObj o = cursor->next().getOwned();
                    {
                        writelock lk( ns );
                        Helpers::upsert( ns , o );
@@ -756,7 +760,11 @@ namespace mongo {
                    BSONObj res;
                    if ( ! conn->runCommand( "admin" , BSON( "_transferMods" << 1 ) , res ) ){
                        state = FAIL;
+                        errmsg = "_transferMods failed: ";
+                        errmsg += res.toString();
                        log( LL_ERROR ) << "_transferMods failed: " << res << endl;
+                        conn.done();
+                        return;
                    }
                    if ( res["size"].number() == 0 )
                        break;
@@ -775,6 +783,7 @@ namespace mongo {
                        log() << "_transferMods failed in STEADY state: " << res << endl;
                        errmsg = res.toString();
                        state = FAIL;
+                        conn.done();
                        return;
                    }

@@ -801,9 +810,10 @@ namespace mongo {
            b.append( "from" , from );
            b.append( "min" , min );
            b.append( "max" , max );
-
+            
            b.append( "state" , stateString() );
-
+            if ( state == FAIL )
+                b.append( "errmsg" , errmsg );
            {
                BSONObjBuilder bb( b.subobjStart( "counts" ) );
                bb.append( "cloned" , numCloned );
--- a/s/d_state.cpp
+++ b/s/d_state.cpp
@@ -603,7 +603,8 @@ namespace mongo {
        BSONObj x = loc.obj().extractFields(_key);
        
        MyMap::const_iterator a = _map.upper_bound( x );
-        a--;
+        if ( a != _map.begin() )
+            a--;
        
        bool good = x.woCompare( a->second.first ) >= 0 && x.woCompare( a->second.second ) < 0;
 #if 0
--- a/s/d_writeback.cpp
+++ b/s/d_writeback.cpp
@@ -52,7 +52,7 @@ namespace mongo {
    class WriteBackCommand : public Command {
    public:
        virtual LockType locktype() const { return NONE; } 
-        virtual bool slaveOk() const { return false; }
+        virtual bool slaveOk() const { return true; }
        virtual bool adminOnly() const { return true; }
        
        WriteBackCommand() : Command( "writebacklisten" ){}
--- a/s/grid.cpp
+++ b/s/grid.cpp
@@ -115,41 +115,107 @@ namespace mongo {
        _allowLocalShard = allow;
    }

-    bool Grid::addShard( string* name , const string& host , long long maxSize , string& errMsg ){
-        // name is optional
+    bool Grid::addShard( string* name , const ConnectionString& servers , long long maxSize , string& errMsg ){
+        // name can be NULL, so privide a dummy one here to avoid testing it elsewhere
        string nameInternal;
        if ( ! name ) {
            name = &nameInternal;
        }

-        // Check whether the host exists and is operative. In order to be accepted as a new shard, that
-        // mongod must not have any database name that exists already in any other shards. If that
-        // test passes, the new shard's databases are going to be entered as non-sharded db's whose
-        // primary is the newly added shard.
+        // Check whether the host (or set) exists and run several sanity checks on this request. 
+        // There are two set of sanity checks: making sure adding this particular shard is consistent
+        // with the replica set state (if it exists) and making sure this shards databases can be 
+        // brought into the grid without conflict.

        vector<string> dbNames;
        try {
-            ScopedDbConnection newShardConn( host );
+            ScopedDbConnection newShardConn( servers );
            newShardConn->getLastError();
            
            if ( newShardConn->type() == ConnectionString::SYNC ){
                newShardConn.done();
-                errMsg = "can't use sync cluster as a shard.  for replica set, have to use <name>/<server1>,<server2>,...";
+                errMsg = "can't use sync cluster as a shard.  for replica set, have to use <setname>/<server1>,<server2>,...";
                return false;
            }

-            // get the shard's local db's listing
-            BSONObj res;
-            bool ok = newShardConn->runCommand( "admin" , BSON( "listDatabases" << 1 ) , res );
+            BSONObj resIsMaster;
+            bool ok =  newShardConn->runCommand( "admin" , BSON( "isMaster" << 1 ) , resIsMaster );
            if ( !ok ){
                ostringstream ss;
-                ss << "failed listing " << host << " databases:" << res;
+                ss << "failed running isMaster: " << resIsMaster;
                errMsg = ss.str();
                newShardConn.done();
                return false;
            }

-            BSONObjIterator i( res["databases"].Obj() );
+            // if the shard has only one host, make sure it is not part of a replica set
+            string setName = resIsMaster["setName"].str();
+            string commandSetName = servers.getSetName();
+            if ( commandSetName.empty() && ! setName.empty() ){
+                ostringstream ss;
+                ss << "host is part of set: " << setName << " use replica set url format <setname>/<server1>,<server2>,....";
+                errMsg = ss.str();
+                newShardConn.done();
+                return false;
+            }
+
+            // if the shard is part of replica set, make sure it is the right one
+            if ( ! commandSetName.empty() && ( commandSetName != setName ) ){
+                ostringstream ss;
+                ss << "host is part of a different set: " << setName;
+                errMsg = ss.str();
+                newShardConn.done();
+                return false;
+            }
+
+            // if the shard is part of a replica set, make sure all the hosts mentioned in 'servers' are part of 
+            // the set. It is fine if not all members of the set are present in 'servers'.
+            bool foundAll = true;
+            string offendingHost;
+            if ( ! commandSetName.empty() ){
+                set<string> hostSet;
+                BSONObjIterator iter( resIsMaster["hosts"].Obj() );
+                while ( iter.more() ){
+                    hostSet.insert( iter.next().String() ); // host:port
+                }
+
+                vector<HostAndPort> hosts = servers.getServers();
+                for ( size_t i = 0 ; i < hosts.size() ; i++ ){
+                    string host = hosts[i].toString(); // host:port
+                    if ( hostSet.find( host ) == hostSet.end() ){
+                        offendingHost = host;
+                        foundAll = false;
+                        break;
+                    }
+                }
+            }
+            if ( ! foundAll ){
+                ostringstream ss;
+                ss << "host " << offendingHost << " does not belong to replica set " << setName;;
+                errMsg = ss.str();
+                newShardConn.done();
+                return false;
+            }
+
+            // shard name defaults to the name of the replica set
+            if ( name->empty() && ! setName.empty() )
+                    *name = setName;
+
+            // In order to be accepted as a new shard, that mongod must not have any database name that exists already 
+            // in any other shards. If that test passes, the new shard's databases are going to be entered as 
+            // non-sharded db's whose primary is the newly added shard.
+
+            BSONObj resListDB;
+            ok = newShardConn->runCommand( "admin" , BSON( "listDatabases" << 1 ) , resListDB );
+            if ( !ok ){
+                ostringstream ss;
+                ss << "failed listing " << servers.toString() << "'s databases:" << resListDB;
+                errMsg = ss.str();
+                newShardConn.done();
+                return false;
+            }
+
+            BSONObjIterator i( resListDB["databases"].Obj() );
            while ( i.more() ){
                BSONObj dbEntry = i.next().Obj();
                const string& dbName = dbEntry["name"].String();
@@ -176,7 +242,7 @@ namespace mongo {
            DBConfigPtr config = getDBConfig( *it , false );
            if ( config.get() != NULL ){
                ostringstream ss;
-                ss << "trying to add shard " << host << " because local database " << *it;
+                ss << "trying to add shard " << servers.toString() << " because local database " << *it;
                ss << " exists in another " << config->getPrimary().toString();
                errMsg = ss.str();
                return false;
@@ -192,7 +258,7 @@ namespace mongo {
        // build the ConfigDB shard document
        BSONObjBuilder b;
        b.append( "_id" , *name );
-        b.append( "host" , host );
+        b.append( "host" , servers.toString() );
        if ( maxSize > 0 ){
            b.append( ShardFields::maxSize.name() , maxSize );
        }
@@ -201,8 +267,8 @@ namespace mongo {
        {
            ScopedDbConnection conn( configServer.getPrimary() );
                
-            // check whether this host:port is not an already a known shard
-            BSONObj old = conn->findOne( ShardNS::shard , BSON( "host" << host ) );
+            // check whether the set of hosts (or single host) is not an already a known shard
+            BSONObj old = conn->findOne( ShardNS::shard , BSON( "host" << servers.toString() ) );
            if ( ! old.isEmpty() ){
                errMsg = "host already used";
                conn.done();
@@ -228,7 +294,7 @@ namespace mongo {
        for ( vector<string>::const_iterator it = dbNames.begin(); it != dbNames.end(); ++it ){
            DBConfigPtr config = getDBConfig( *it , true , *name );
            if ( ! config ){
-                log() << "adding shard " << host << " even though could not add database " << *it << endl; 
+                log() << "adding shard " << servers << " even though could not add database " << *it << endl; 
            }
        }

--- a/s/grid.h
+++ b/s/grid.h
@@ -57,18 +57,17 @@ namespace mongo {
        /**
         *
         * addShard will create a new shard in the grid. It expects a mongod process to be runing
-         * on the provided address.
-         * TODO - add the mongod's databases to the grid
+         * on the provided address. Adding a shard that is a replica set is supported. 
         *
         * @param name is an optional string with the name of the shard. if ommited, grid will
         * generate one and update the parameter.
-         * @param host is the complete address of the machine where the shard will be
+         * @param servers is the connection string of the shard being added
         * @param maxSize is the optional space quota in bytes. Zeros means there's no limitation to
         * space usage
         * @param errMsg is the error description in case the operation failed. 
         * @return true if shard was successfully added.
         */
-        bool addShard( string* name , const string& host , long long maxSize , string& errMsg );
+        bool addShard( string* name , const ConnectionString& servers , long long maxSize , string& errMsg );

        /**
         * @return true if the config database knows about a host 'name'
--- a/s/request.cpp
+++ b/s/request.cpp
@@ -123,7 +123,8 @@ namespace mongo {
                log() << staleConfig.what() << " attempt: " << attempt << endl;
                uassert( 10195 ,  "too many attempts to update config, failing" , attempt < 5 );
                ShardConnection::checkMyConnectionVersions( getns() );
-                sleepsecs( attempt );
+                if (!staleConfig.justConnection() )
+                    sleepsecs( attempt );
                reset( ! staleConfig.justConnection() );
                _d.markReset();
                process( attempt + 1 );
--- a/s/server.cpp
+++ b/s/server.cpp
@@ -63,10 +63,7 @@ namespace mongo {

    class ShardingConnectionHook : public DBConnectionHook {
    public:
-        virtual void onCreate( DBClientBase * conn ){
-            if ( conn->type() != ConnectionString::SYNC )
-                conn->simpleCommand( "admin" , 0 , "switchtoclienterrors" );
-        }
+
        virtual void onHandedOut( DBClientBase * conn ){
            ClientInfo::get()->addShard( conn->getServerAddress() );
        }
@@ -92,10 +89,12 @@ namespace mongo {
                r.process();
            }
            catch ( DBException& e ){
+                log() << "DBException in process: " << e.what() << endl;
+                
                le->raiseError( e.getCode() , e.what() );
                
                m.header()->id = r.id();
-                log() << "UserException: " << e.what() << endl;
+                
                if ( r.expectResponse() ){
                    BSONObj err = BSON( "$err" << e.what() << "code" << e.getCode() );
                    replyToQuery( ResultFlag_ErrSet, p , m , err );
--- a/s/shard.cpp
+++ b/s/shard.cpp
@@ -76,6 +76,12 @@ namespace mongo {
                Shard s( name , host , maxSize , isDraining );
                _lookup[name] = s;
                _lookup[host] = s;
+
+                // add rs name to lookup (if it exists)
+                size_t pos;
+                if ((pos = host.find('/', 0)) != string::npos) {
+                    _lookup[host.substr(0, pos)] = s;
+                }
            }

        }
@@ -90,6 +96,13 @@ namespace mongo {
            {
                scoped_lock lk( _mutex );
                map<string,Shard>::iterator i = _lookup.find( ident );
+
+                // if normal find didn't find anything, try to find by rs name
+                size_t pos;
+                if ( i == _lookup.end() && (pos = ident.find('/', 0)) != string::npos) {
+                    i = _lookup.find( ident.substr(0, pos) );
+                }
+
                if ( i != _lookup.end() )
                    return i->second;
            }
--- a/s/shardconnection.cpp
+++ b/s/shardconnection.cpp
@@ -26,6 +26,7 @@ namespace mongo {
    
    /**
     * holds all the actual db connections for a client to various servers
+     * 1 pre thread, so don't have to worry about thread safety
     */
    class ClientConnections : boost::noncopyable {
    public:
@@ -70,7 +71,8 @@ namespace mongo {
            _hosts.clear();
        }
        
-        DBClientBase * get( const string& addr ){
+        DBClientBase * get( const string& addr , const string& ns ){
+            _check( ns );
            scoped_lock lk( _mutex );
            Status* &s = _hosts[addr];
            if ( ! s )
@@ -152,9 +154,16 @@ namespace mongo {
            }
        }
        
+        void _check( const string& ns ){
+            if ( ns.size() == 0 || _seenNS.count( ns ) )
+                return;
+            _seenNS.insert( ns );
+            checkVersions( ns );
+        }
+
        map<string,Status*> _hosts;
        mongo::mutex _mutex;
-
+        set<string> _seenNS;
        // -----
        
        static thread_specific_ptr<ClientConnections> _perThread;
@@ -188,7 +197,7 @@ namespace mongo {
    
    void ShardConnection::_init(){
        assert( _addr.size() );
-        _conn = ClientConnections::get()->get( _addr );
+        _conn = ClientConnections::get()->get( _addr , _ns );
        _finishedInit = false;
    }

--- a/s/util.h
+++ b/s/util.h
@@ -58,11 +58,19 @@ namespace mongo {
            }
        }

-        ShardChunkVersion incMajor() const {
-            return ShardChunkVersion( _major + 1 , 0 );
+        void inc( bool major ){
+            if ( major )
+                incMajor();
+            else
+                incMinor();
        }

-        void operator++(){
+        void incMajor() {
+            _major++;
+            _minor = 0;
+        }
+        
+        void incMinor() {
            _minor++;
        }

@@ -79,8 +87,9 @@ namespace mongo {
            ss << _major << "|" << _minor; 
            return ss.str(); 
        }
+        
        operator unsigned long long() const { return _combined; }
-
+        
        ShardChunkVersion& operator=( const BSONElement& elem ){
            switch ( elem.type() ){
            case Timestamp:
--- a/scripting/sm_db.cpp
+++ b/scripting/sm_db.cpp
@@ -156,7 +156,7 @@ namespace mongo {
    JSBool mongo_external_constructor( JSContext *cx, JSObject *obj, uintN argc, jsval *argv, jsval *rval ){
        Convertor c( cx );
        
-        uassert( 10238 ,  "0 or 1 args to Mongo" , argc <= 1 );
+        smuassert( cx ,  "0 or 1 args to Mongo" , argc <= 1 );
        
        string host = "127.0.0.1";
        if ( argc > 0 )
@@ -207,9 +207,9 @@ namespace mongo {
     };

    JSBool mongo_find(JSContext *cx, JSObject *obj, uintN argc, jsval *argv, jsval *rval){
-        uassert( 10240 ,  "mongo_find neesd 6 args" , argc == 6 );
+        smuassert( cx , "mongo_find needs 6 args" , argc == 6 );
        shared_ptr< DBClientWithCommands > * connHolder = (shared_ptr< DBClientWithCommands >*)JS_GetPrivate( cx , obj );
-        uassert( 10241 ,  "no connection!" , connHolder && connHolder->get() );
+        smuassert( cx ,  "no connection!" , connHolder && connHolder->get() );
        DBClientWithCommands *conn = connHolder->get();
                      
        Convertor c( cx );
--- a/shell/query.js
+++ b/shell/query.js
@@ -243,6 +243,20 @@ DBQuery.prototype.explain = function (verbose) {
    if (!verbose) {
        delete e.allPlans;
        delete e.oldPlan;
+        if (e.shards){
+            for (var key in e.shards){
+                var s = e.shards[key];
+                if(s.length === undefined){
+                    delete s.allPlans;
+                    delete s.oldPlan;
+                } else {
+                    for (var i=0; i < s.length; i++){
+                        delete s[i].allPlans;
+                        delete s[i].oldPlan;
+                    }
+                }
+            }
+        }
    }
    return e;
 }
--- a/shell/servers.js
+++ b/shell/servers.js
@@ -62,6 +62,13 @@ startMongodTest = function (port, dirname, restart, extraOptions ) {
        f = startMongodNoReset;
    if (!dirname)
        dirname = "" + port; // e.g., data/db/27000
+
+    var useHostname = false;
+    if (extraOptions) {
+         useHostname = extraOptions.useHostname;
+         delete extraOptions.useHostname;
+    }
+
    
    var options = 
        {
@@ -78,7 +85,7 @@ startMongodTest = function (port, dirname, restart, extraOptions ) {
    
    var conn = f.apply(null, [ options ] );

-    conn.name = "localhost:" + port;
+    conn.name = (useHostname ? getHostName() : "localhost") + ":" + port;
    return conn;
 }

@@ -145,6 +152,10 @@ myPort = function() {
        return 27017;
 }

+/**
+ * otherParams can be:
+ * * useHostname to use the hostname (instead of localhost)
+ */
 ShardingTest = function( testName , numShards , verboseLevel , numMongos , otherParams ){
    this._testName = testName;

@@ -155,7 +166,10 @@ ShardingTest = function( testName , numShards , verboseLevel , numMongos , other
    if ( otherParams.sync && numShards < 3 )
        throw "if you want sync, you need at least 3 servers";

-    var localhost = "localhost";
+    var localhost = otherParams.useHostname ? getHostName() : "localhost";
+
+    this._alldbpaths = []
+

    if ( otherParams.rs ){
        localhost = getHostName();
@@ -165,6 +179,11 @@ ShardingTest = function( testName , numShards , verboseLevel , numMongos , other
            var rs = new ReplSetTest( { name : testName + "-rs" + i , nodes : 3 , startPort : 31100 + ( i * 100 ) } );
            this._rs[i] = { test : rs , nodes : rs.startSet( { oplogSize:40 } ) , url : rs.getURL() };
            rs.initiate();
+            
+        }
+
+        for ( var i=0; i<numShards; i++){
+            var rs = this._rs[i].test;
            rs.getMaster().getDB( "admin" ).foo.save( { x : 1 } )
            rs.awaitReplication();
            this._connections.push( new Mongo( rs.getURL() ) );
@@ -173,6 +192,7 @@ ShardingTest = function( testName , numShards , verboseLevel , numMongos , other
        this._configServers = []
        for ( var i=0; i<3; i++ ){
            var conn = startMongodTest( 30000 + i , testName + "-config" + i );
+            this._alldbpaths.push( testName + "-config" + i )
            this._configServers.push( conn );
        }
        
@@ -182,17 +202,18 @@ ShardingTest = function( testName , numShards , verboseLevel , numMongos , other
    }
    else {
        for ( var i=0; i<numShards; i++){
-            var conn = startMongodTest( 30000 + i , testName + i );
+            var conn = startMongodTest( 30000 + i , testName + i, 0, {useHostname : otherParams.useHostname} );
+            this._alldbpaths.push( testName +i )
            this._connections.push( conn );
        }
        
        if ( otherParams.sync ){
-            this._configDB = "localhost:30000,localhost:30001,localhost:30002";
+            this._configDB = localhost+":30000,"+localhost+":30001,"+localhost+":30002";
            this._configConnection = new Mongo( this._configDB );
            this._configConnection.getDB( "config" ).settings.insert( { _id : "chunksize" , value : otherParams.chunksize || 50 } );        
        }
        else {
-            this._configDB = "localhost:30000";
+            this._configDB = localhost + ":30000";
            this._connections[0].getDB( "config" ).settings.insert( { _id : "chunksize" , value : otherParams.chunksize || 50 } );
        }
    }
@@ -201,6 +222,7 @@ ShardingTest = function( testName , numShards , verboseLevel , numMongos , other
    var startMongosPort = 31000;
    for ( var i=0; i<(numMongos||1); i++ ){
        var myPort =  startMongosPort - i;
+        print("config: "+this._configDB);
        var conn = startMongos( { port : startMongosPort - i , v : verboseLevel || 0 , configdb : this._configDB }  );
        conn.name = localhost + ":" + myPort;
        this._mongos.push( conn );
@@ -316,6 +338,11 @@ ShardingTest.prototype.stop = function(){
            this._rs[i].test.stopSet( 15 );
        }
    }
+    if ( this._alldbpaths ){
+        for( i=0; i<this._alldbpaths.length; i++ ){
+            resetDbpath( "/data/db/" + this._alldbpaths[i] );
+        }
+    }

    print('*** ' + this._testName + " completed successfully ***");
 }
@@ -966,6 +993,8 @@ ReplSetTest = function( opts ){
    this.name  = opts.name || "testReplSet";
    this.host  = opts.host || getHostName();
    this.numNodes = opts.nodes || 0;
+    this.useSeedList = opts.useSeedList || false;
+
    this.bridged = opts.bridged || false;
    this.ports = [];

@@ -1024,6 +1053,10 @@ ReplSetTest.prototype.getPort = function( n ){
 ReplSetTest.prototype.getPath = function( n ){
    var p = "/data/db/" + this.name + "-";
    p += n.toString();
+    if ( ! this._alldbpaths )
+        this._alldbpaths = [ p ];
+    else
+        this._alldbpaths.push( p );
    return p;
 }

@@ -1326,9 +1359,15 @@ ReplSetTest.prototype.stop = function( n , signal ){
    return stopMongod( port , signal || 15 );
 }

-ReplSetTest.prototype.stopSet = function( signal ) {
+ReplSetTest.prototype.stopSet = function( signal , forRestart ) {
    for(i=0; i < this.ports.length; i++) {
        this.stop( i, signal );
    }
+    if ( ! forRestart && this._alldbpaths ){
+        for( i=0; i<this._alldbpaths.length; i++ ){
+            resetDbpath( this._alldbpaths[i] );
+        }
+    }
+
    print('*** Shut down repl set - test worked ****' )
 }
--- a/util/background.cpp
+++ b/util/background.cpp
@@ -65,8 +65,7 @@ namespace mongo {
    }

    bool BackgroundJob::wait(int msMax, unsigned maxsleep) {
-        assert( state != NotStarted );
-        unsigned ms = 0;
+        unsigned ms = 1;
        Date_t start = jsTime();
        while ( state != Done ) {
            sleepmillis(ms);
@@ -84,7 +83,7 @@ namespace mongo {

    /* wait for several jobs to finish. */
    void BackgroundJob::wait(list<BackgroundJob*>& L, unsigned maxsleep) {
-        unsigned ms = 0;
+        unsigned ms = 1;
        {
            x:
            sleepmillis(ms);
--- a/util/concurrency/msg.h
+++ b/util/concurrency/msg.h
@@ -27,13 +27,13 @@ namespace mongo {

        typedef boost::function<void()> lam;

-        /** typical usage is: task::fork( serverPtr ); */
+        /** typical usage is: task::fork( new Server("threadname") ); */
        class Server : public Task { 
        public:
            /** send a message to the port */
            void send(lam);

-            Server(string name) : _name(name) { }
+            Server(string name) : _name(name), rq(false) { }
            virtual ~Server() { }

            /** send message but block until function completes */
@@ -42,9 +42,8 @@ namespace mongo {
            void requeue() { rq = true; }

        protected:
-            /* this needn't be abstract; i left it that way for now so i remember 
-               to call Client::initThread() when using in mongo... */
-            virtual void starting() = 0;
+            /* REMINDER : for use in mongod, you will want to have this call Client::initThread(). */
+            virtual void starting() { }

        private:
            virtual bool initClient() { return true; }
--- a/util/concurrency/task.cpp
+++ b/util/concurrency/task.cpp
@@ -117,7 +117,6 @@ namespace mongo {

        void Server::doWork() { 
            starting();
-            rq = false;
            while( 1 ) { 
                lam f;
                try {
@@ -134,7 +133,10 @@ namespace mongo {
                    f();
                    if( rq ) {
                        rq = false;
-                        send(f);
+                        {
+                            boost::mutex::scoped_lock lk(m);
+                            d.push_back(f);
+                        }
                    }
                } catch(std::exception& e) { 
                    log() << "Server::doWork() exception " << e.what() << endl;
@@ -143,6 +145,27 @@ namespace mongo {
                }
            }
        }
-        
+
+        static Server *s;
+        static void abc(int i) { 
+            cout << "Hello " << i << endl;
+            s->requeue();
+        }
+        class TaskUnitTest : public mongo::UnitTest {
+        public:
+            virtual void run() { 
+                lam f = boost::bind(abc, 3);
+                //f();
+
+                s = new Server("unittest");
+                fork(s);
+                s->send(f);
+
+                sleepsecs(30);
+                cout <<" done" << endl;
+
+            }
+        }; // not running. taskunittest;
+
    }
 }
--- a/util/hostandport.h
+++ b/util/hostandport.h
@@ -74,6 +74,7 @@ namespace mongo {
        string host() const { return _host; }

        int port() const { return _port >= 0 ? _port : CmdLine::DefaultDBPort; }
+        bool hasPort() const { return _port >= 0; }
        void setPort( int port ) { _port = port; }

    private:
--- a/util/log.h
+++ b/util/log.h
@@ -158,15 +158,7 @@ namespace mongo {
        static vector<Tee*> * globalTees;
    public:

-        static void logLockless( const StringData& s ){
-            if ( doneSetup == 1717 ){
-                fwrite( s.data() , s.size() , 1 , logfile );
-                fflush( logfile );
-            }
-            else {
-                cout << s.data() << endl;
-            }
-        }
+        inline static void logLockless( const StringData& s );
        
        static void setLogFile(FILE* f){
            scoped_lock lk(mutex);
@@ -177,52 +169,9 @@ namespace mongo {
            return 1717;
        }

-        void flush(Tee *t = 0) {
-            // this ensures things are sane
-            if ( doneSetup == 1717 ) {
-                string msg = ss.str();
-                string threadName = getThreadName();
-                const char * type = logLevelToString(logLevel);
-                
-                int spaceNeeded = msg.size() + 64 + threadName.size();
-                int bufSize = 128;
-                while ( bufSize < spaceNeeded )
-                    bufSize += 128;
-                
-                BufBuilder b(bufSize);
-                time_t_to_String( time(0) , b.grow(20) );
-                if (!threadName.empty()){
-                    b.appendChar( '[' );
-                    b.appendStr( threadName , false );
-                    b.appendChar( ']' );
-                    b.appendChar( ' ' );
-                }
-                if ( type[0] ){
-                    b.appendStr( type , false );
-                    b.appendStr( ": " , false );
-                }
-                b.appendStr( msg );
-                
-                string out( b.buf() , b.len() - 1);
-
-                scoped_lock lk(mutex);
-                
-                if( t ) t->write(logLevel,out);
-                if ( globalTees ){
-                    for ( unsigned i=0; i<globalTees->size(); i++ )
-                        (*globalTees)[i]->write(logLevel,out);
-                }
-                
-#ifndef _WIN32
-                //syslog( LOG_INFO , "%s" , cc );
-#endif
-                fwrite(out.data(), out.size(), 1, logfile);
-                fflush(logfile);
-            }
-            _init();
-        }
+        inline void flush(Tee *t = 0);
        
-        Nullstream& setLogLevel(LogLevel l){
+        inline Nullstream& setLogLevel(LogLevel l){
            logLevel = l;
            return *this;
        }
@@ -416,6 +365,69 @@ namespace mongo {
    /** output the error # and error message with prefix.  
        handy for use as parm in uassert/massert.
        */
-    string errnoWithPrefix( const char * prefix = 0 );
+    string errnoWithPrefix( const char * prefix );
+
+    void Logstream::logLockless( const StringData& s ){
+        if ( doneSetup == 1717 ){
+            if(fwrite(s.data(), s.size(), 1, logfile)){
+                fflush(logfile);
+            }else{
+                int x = errno;
+                cout << "Failed to write to logfile: " << errnoWithDescription(x) << ": " << out << endl;
+            }
+        }
+        else {
+            cout << s.data() << endl;
+        }
+    }
+
+    void Logstream::flush(Tee *t) {
+        // this ensures things are sane
+        if ( doneSetup == 1717 ) {
+            string msg = ss.str();
+            string threadName = getThreadName();
+            const char * type = logLevelToString(logLevel);
+
+            int spaceNeeded = msg.size() + 64 + threadName.size();
+            int bufSize = 128;
+            while ( bufSize < spaceNeeded )
+                bufSize += 128;
+
+            BufBuilder b(bufSize);
+            time_t_to_String( time(0) , b.grow(20) );
+            if (!threadName.empty()){
+                b.appendChar( '[' );
+                b.appendStr( threadName , false );
+                b.appendChar( ']' );
+                b.appendChar( ' ' );
+            }
+            if ( type[0] ){
+                b.appendStr( type , false );
+                b.appendStr( ": " , false );
+            }
+            b.appendStr( msg );
+
+            string out( b.buf() , b.len() - 1);
+
+            scoped_lock lk(mutex);
+
+            if( t ) t->write(logLevel,out);
+            if ( globalTees ){
+                for ( unsigned i=0; i<globalTees->size(); i++ )
+                    (*globalTees)[i]->write(logLevel,out);
+            }
+
+#ifndef _WIN32
+            //syslog( LOG_INFO , "%s" , cc );
+#endif
+            if(fwrite(out.data(), out.size(), 1, logfile)){
+                fflush(logfile);
+            }else{
+                int x = errno;
+                cout << "Failed to write to logfile: " << errnoWithDescription(x) << ": " << out << endl;
+            }
+        }
+        _init();
+    }

 } // namespace mongo
--- a/util/message.cpp
+++ b/util/message.cpp
@@ -168,7 +168,11 @@ namespace mongo {
            const int ret = select(maxfd+1, fds, NULL, NULL, &maxSelectTime);
            
            if (ret == 0){
+#if defined(__linux__)
                _elapsedTime += ( 10000 - maxSelectTime.tv_usec ) / 1000;
+#else
+                _elapsedTime += 10;
+#endif
                continue;
            }
            _elapsedTime += ret; // assume 1ms to grab connection. very rough
--- a/util/version.cpp
+++ b/util/version.cpp
@@ -14,7 +14,7 @@ namespace mongo {
    // mongo processes version support
    //

-    const char versionString[] = "1.5.9-pre-";
+    const char versionString[] = "1.6.2";

    string mongodVersion() {
        stringstream ss;
Author	SHA1	Message	Date
Eliot Horowitz	aef371ecf5	BUMP 1.6.2	2010-09-01 08:50:38 -04:00
Kristina Chodorow	5b66c47327	fix sharding tests	2010-08-31 23:41:17 -04:00
Alberto Lerner	0260ce050b	SERVER-1607 more replica set checks before adding shard	2010-08-31 21:21:18 -04:00
Alberto Lerner	3c1fa57776	SERVER-1607 replica set checks before adding a shard	2010-08-31 21:21:06 -04:00
Kristina Chodorow	19a0f996fa	tests SERVER-1665	2010-08-31 21:20:58 -04:00
Alberto Lerner	0fda3ce40f	SERVER-1607 expose replica set of a ConnectionString	2010-08-31 21:20:16 -04:00
Alberto Lerner	4a9135c3b6	SERVER-1607 isMaster returns the name of the replica set, if appropriate	2010-08-31 21:17:31 -04:00
Eliot Horowitz	0af2db40cf	fix test for v1.6	2010-08-31 21:16:07 -04:00
Dwight	f93b0d654f	move tests that close the database to slownightly so they dont interfere with other tests running at the same time	2010-08-31 20:18:05 -04:00
Alberto Lerner	9ff3ddc944	SERVER-1611 allow localhost in a replica-set-based shard	2010-08-31 17:38:05 -04:00
Eliot Horowitz	cd47c92322	fix segfault on repl set slave before inited SERVER-1656 1.6 version	2010-08-31 17:36:10 -04:00
Eliot Horowitz	2b869d6dfa	no size_t append in 1.6	2010-08-31 17:33:58 -04:00
Dwight	c9feeb823d	fix bugs with query yielding 162	2010-08-31 17:28:08 -04:00
Dwight	38da020af0	elim ambiguous call warning/error	2010-08-31 17:27:28 -04:00
dwight	79b5c8023b	const is safer	2010-08-31 17:27:14 -04:00
dwight	baa4cb9c04	clientcursor finish last commit	2010-08-31 17:27:04 -04:00
dwight	66befc517c	clean up diags clientcursor after changes	2010-08-31 17:26:56 -04:00
dwight	e070747d94	fix an old but serious bug with cursors 162	2010-08-31 17:26:45 -04:00
Eliot Horowitz	921dbd018b	major delete issue with multiple dbs	2010-08-31 17:26:42 -04:00
Eliot Horowitz	964b3436ef	remove warning when advancing a cursor to the end	2010-08-31 17:25:26 -04:00
Eliot Horowitz	0840daf31a	track number of timed out cursors and display in cursorinfo and serverStatus	2010-08-31 17:25:16 -04:00
Dwight	a3dd77e7ec	lower risk of collection name collisions	2010-08-31 17:22:30 -04:00
Eliot Horowitz	4a65b51c57	assert when would inf. loop, but probably caused by something else, still looking	2010-08-31 17:21:23 -04:00
Eliot Horowitz	bdfefa42c5	add if debug is on to buildinfo	2010-08-31 17:17:52 -04:00
Eliot Horowitz	625b4c4378	fix docs for pidfilepath SERVER-1706	2010-08-31 17:16:23 -04:00
Eliot Horowitz	4099bc52cc	mongo::Query instead of Query for define	2010-08-31 17:15:44 -04:00
Eliot Horowitz	5c6f0dcfbd	better debugging for drop errors	2010-08-31 17:15:31 -04:00
Eliot Horowitz	b620e3327e	don't log create collection if on replication side SERVER-1707	2010-08-31 11:18:51 -04:00
Kristina Chodorow	92d5eb42af	use replica set name for shard id SERVER-1665	2010-08-28 12:01:22 -04:00
Eliot Horowitz	aaae92dedb	make web console get query in a thread safe way SERVER-1692	2010-08-26 13:54:32 -04:00
Eliot Horowitz	0e78196fd5	use smuassert instead of uassert SERVER-1661	2010-08-23 13:37:13 -04:00
Eliot Horowitz	6122c120c3	sharded m/r supports scope SERVER-1663	2010-08-23 13:20:50 -04:00
Eliot Horowitz	7b02bf761d	fix m/r sort SERVER-1652	2010-08-21 21:24:09 -04:00
Eliot Horowitz	00e9040236	sayPiggyBack for replica set connection SERVER-1647	2010-08-20 14:12:40 -04:00
Eliot Horowitz	f0d0aad23d	fix change log	2010-08-20 11:53:19 -04:00
Eliot Horowitz	f6aad0935e	remove mechanism to put mongos -> mongod connections in last error id mode	2010-08-19 10:14:01 -04:00
Eliot Horowitz	7093e6e585	1.6.2 prep	2010-08-17 16:02:25 -04:00
Eliot Horowitz	c5f5f9a4f3	stabilize test	2010-08-17 12:28:19 -04:00
Alberto Lerner	19b587d925	Depend less an a time estimate.	2010-08-17 11:49:07 -04:00
Eliot Horowitz	31d81b823d	BUMP 1.6.1	2010-08-17 10:04:43 -04:00
Eliot Horowitz	e280c561f1	make sharding/rs test faster	2010-08-17 09:57:17 -04:00
Eliot Horowitz	1a30dad111	turn off test that is different on windows	2010-08-17 09:34:29 -04:00
Eliot Horowitz	3770916e05	SERVER-1626 initial sync bug fix Conflicts: db/repl.cpp	2010-08-16 17:07:08 -04:00
Eliot Horowitz	67159d2583	BackgroundJob::wait bug SERVER-1628 Conflicts: util/background.cpp	2010-08-16 17:04:22 -04:00
dwight	81da9cda7f	faster	2010-08-16 17:03:38 -04:00
dwight	415c264123	SERVER-1614	2010-08-16 17:02:33 -04:00
Dwight	4454a8a0a0	fix help title	2010-08-16 16:57:23 -04:00
Eliot Horowitz	d1c3edcf33	writeback is ok on slave	2010-08-16 15:48:08 -04:00
Eliot Horowitz	1948efc6a4	cursor paranoia	2010-08-16 14:30:11 -04:00
Eliot Horowitz	c3855123b5	even more migrate debugging	2010-08-16 14:29:59 -04:00
Eliot Horowitz	2976396903	return connection to pool	2010-08-16 14:29:48 -04:00
Eliot Horowitz	aa5d314ae9	compile	2010-08-16 14:29:39 -04:00
Eliot Horowitz	d3b427b96f	better transfer error logging	2010-08-16 14:29:29 -04:00
Eliot Horowitz	3de268fcae	change error reporting order for robustness	2010-08-16 14:28:53 -04:00
Eliot Horowitz	54239a3676	fix select() timing for os x SERVER-1596	2010-08-16 14:28:12 -04:00
Eliot Horowitz	1d7dfe2868	try to fix select timing on windows SERVER-1596	2010-08-16 14:28:02 -04:00
Eliot Horowitz	f1f53ef3c1	possible chunk matcher fix	2010-08-16 14:27:43 -04:00
Eliot Horowitz	356eeb5a62	make BSONElement::toString 20x faster for double	2010-08-16 14:27:29 -04:00
Eliot Horowitz	2800a96e56	clean up chunk version handling - prep for SERVER-1473	2010-08-16 14:26:12 -04:00
Mathias Stearn	0c8f696e79	be smarter about sleeping	2010-08-16 14:25:59 -04:00
Alberto Lerner	b58d62a194	Bug fix: logChange() does't throw (can be called from destructors	2010-08-16 14:16:58 -04:00
Eliot Horowitz	2ba8ac25f1	CS-112 initial sync problem with traditional replication Conflicts: db/repl.cpp	2010-08-12 16:12:02 -04:00
Mathias Stearn	edb14c522b	move log flush code to after errnoWithDescription	2010-08-11 14:43:26 -04:00
Mathias Stearn	af35eebfb6	check return from fwrite SERVER-1578	2010-08-11 14:43:16 -04:00
Mathias Stearn	edffe867a5	distinct should yield even if non-matching. SERVER-1597	2010-08-09 14:45:12 -04:00
Mathias Stearn	0e8ef6bca0	Yield in distinct SERVER-1597	2010-08-09 14:45:03 -04:00
Mathias Stearn	b59584f3dd	Support large files in GridFS SERVER-1408	2010-08-09 13:38:43 -04:00
Eliot Horowitz	3cc86177ce	fix getMore with sharding+rs in edge case SERVER-1584	2010-08-06 13:25:52 -04:00
Alberto Lerner	dcfab78c4b	DistributedLock::unlock() handles partition from config DB	2010-08-05 13:41:01 -04:00
Eliot Horowitz	ba166df412	post 1.6.0	2010-08-05 11:34:06 -04:00
Eliot Horowitz	2c7f164b65	BUMP 1.6.0	2010-08-05 09:32:06 -04:00
Alberto Lerner	53d1c8bbad	Bug fix: DefaultDBPort is perfectly valid when transitioning from a non-sharded mongod	2010-08-05 00:16:50 -04:00
Eliot Horowitz	e4fcc800f7	fix serverStatus for replica sets SERVER-1570	2010-08-04 16:03:32 -04:00
Eliot Horowitz	7abf0580ca	when restarting repl set don't blow away directories Conflicts: shell/servers.js	2010-08-04 16:00:06 -04:00
Eliot Horowitz	ae1e0804da	fix upsert with $atomic SERVER-1568	2010-08-04 15:15:30 -04:00
Mathias Stearn	50b7bbe934	Low-verbosity sharded explain() SERVER-1562	2010-08-04 13:30:47 -04:00
Eliot Horowitz	fd26b536e9	clean up data files after sharding/replset tests	2010-08-04 12:33:13 -04:00
Dwight	6f6aa20e92	trying to fix SERVER-1483	2010-08-04 12:13:38 -04:00
Dwight	55937a8f3e	sort members in replSetGetStatus output for easy reading	2010-08-04 12:11:18 -04:00
Richard Kreuter	8947d06986	Remove C++ driver from RPM specfile.	2010-08-04 11:54:00 -04:00
Dwight	8c095b2cb2	make error msgs better	2010-08-04 11:53:56 -04:00
Dwight	2e6aeb0229	extraneous logging	2010-08-04 11:53:52 -04:00
Dwight	cb661f73fd	allow a member down during a config change	2010-08-04 11:53:44 -04:00
Eliot Horowitz	63f894763d	fix sorting if sort field not included in projection SERVER-1561	2010-08-04 11:53:36 -04:00
Eliot Horowitz	2c3d88336e	allow get last errors blocking on w to be killed	2010-08-04 10:20:16 -04:00
Eliot Horowitz	c542dbd78a	when access namespace for the first time, prep all connections SERVER-1560	2010-08-04 10:17:17 -04:00
Eliot Horowitz	b5ce278c18	fix type SERVER-1563	2010-08-04 09:39:16 -04:00
Eliot Horowitz	a432ac592e	1.6.0-pre- marker	2010-08-03 22:18:44 -04:00