Open a journal file on startup 180

This reverts commit 049436fd5b.

client/connpool.cpp

@@ -37,7 +37,12 @@ namespace mongo {
     }
 
     void PoolForHost::done( DBClientBase * c ) {
-        _pool.push(c);
+        if ( _pool.size() >= _maxPerHost ) {
+            delete c;
+        }
+        else {
+            _pool.push(c);
+        }
     }
 
     DBClientBase * PoolForHost::get() {
@@ -86,6 +91,8 @@ namespace mongo {
         _created++;
     }
 
+    unsigned PoolForHost::_maxPerHost = 50;
+
     // ------ DBConnectionPool ------
 
     DBConnectionPool pool;

client/connpool.h

@@ -57,6 +57,9 @@ namespace mongo {
         void done( DBClientBase * c );
 
         void flush();
+
+        static void setMaxPerHost( unsigned max ) { _maxPerHost = max; }
+        static unsigned getMaxPerHost() { return _maxPerHost; }
     private:
 
         struct StoredConnection {
@@ -71,6 +74,8 @@ namespace mongo {
         std::stack<StoredConnection> _pool;
         long long _created;
         ConnectionString::ConnectionType _type;
+
+        static unsigned _maxPerHost;
     };
 
     class DBConnectionHook {

client/dbclient_rs.cpp

@@ -183,18 +183,16 @@ namespace mongo {
 
 
     HostAndPort ReplicaSetMonitor::getMaster() {
-        bool good = false;
         {
             scoped_lock lk( _lock );
-            good = _master >= 0 && _nodes[_master].ok;
+            if ( _master >= 0 && _nodes[_master].ok )
+                return _nodes[_master].addr;
         }
-
-        if ( ! good )
-            _check();
-
-        uassert( 10009 , str::stream() << "ReplicaSetMonitor no master found for set: " << _name , _master >= 0 );
-
+        
+        _check();
+        
         scoped_lock lk( _lock );
+        uassert( 10009 , str::stream() << "ReplicaSetMonitor no master found for set: " << _name , _master >= 0 );
         return _nodes[_master].addr;
     }
     

db/btree.cpp

@@ -1114,16 +1114,17 @@ namespace mongo {
 
     /** remove a key from the index */
     bool BtreeBucket::unindex(const DiskLoc thisLoc, IndexDetails& id, const BSONObj& key, const DiskLoc recordLoc ) const {
-        if ( key.objsize() > KeyMax ) {
-            OCCASIONALLY problem() << "unindex: key too large to index, skipping " << id.indexNamespace() << /* ' ' << key.toString() << */ endl;
-            return false;
-        }
-
         int pos;
         bool found;
         DiskLoc loc = locate(id, thisLoc, key, Ordering::make(id.keyPattern()), pos, found, recordLoc, 1);
         if ( found ) {
+
+            if ( key.objsize() > KeyMax ) {
+                OCCASIONALLY problem() << "unindex: key too large to index but was found for " << id.indexNamespace() << " reIndex suggested" << endl;
+            }
+            
             loc.btreemod()->delKeyAtPos(loc, id, pos, Ordering::make(id.keyPattern()));
+            
             return true;
         }
         return false;
@@ -1740,13 +1741,18 @@ namespace mongo {
     }
 
     void BtreeBuilder::mayCommitProgressDurably() {
-        RARELY {
-            getDur().commitIfNeeded();
+        if ( getDur().commitIfNeeded() ) {
             b = cur.btreemod();
         }
     }
 
     void BtreeBuilder::addKey(BSONObj& key, DiskLoc loc) {
+        if ( key.objsize() > KeyMax ) {
+            problem() << "Btree::insert: key too large to index, skipping " << idx.indexNamespace() 
+                      << ' ' << key.objsize() << ' ' << key.toString() << endl;
+            return;
+        }
+
         if( !dupsAllowed ) {
             if( n > 0 ) {
                 int cmp = keyLast.woCompare(key, order);
@@ -1760,15 +1766,9 @@ namespace mongo {
         }
 
         if ( ! b->_pushBack(loc, key, ordering, DiskLoc()) ) {
-            // no room
-            if ( key.objsize() > KeyMax ) {
-                problem() << "Btree::insert: key too large to index, skipping " << idx.indexNamespace() << ' ' << key.objsize() << ' ' << key.toString() << endl;
-            }
-            else {
-                // bucket was full
-                newBucket();
-                b->pushBack(loc, key, ordering, DiskLoc());
-            }
+            // bucket was full
+            newBucket();
+            b->pushBack(loc, key, ordering, DiskLoc());
         }
         n++;
         mayCommitProgressDurably();
@@ -1790,6 +1790,11 @@ namespace mongo {
 
             DiskLoc xloc = loc;
             while( !xloc.isNull() ) {
+                if ( getDur().commitIfNeeded() ) {
+                    b = cur.btreemod();
+                    up = upLoc.btreemod();
+                }
+
                 BtreeBucket *x = xloc.btreemod();
                 BSONObj k;
                 DiskLoc r;
@@ -1833,18 +1838,21 @@ namespace mongo {
     }
 
     BtreeBuilder::~BtreeBuilder() {
-        if( !committed ) {
-            log(2) << "Rolling back partially built index space" << endl;
-            DiskLoc x = first;
-            while( !x.isNull() ) {
-                DiskLoc next = x.btree()->tempNext();
-                string ns = idx.indexNamespace();
-                theDataFileMgr._deleteRecord(nsdetails(ns.c_str()), ns.c_str(), x.rec(), x);
-                x = next;
+        DESTRUCTOR_GUARD(
+            if( !committed ) {
+                log(2) << "Rolling back partially built index space" << endl;
+                DiskLoc x = first;
+                while( !x.isNull() ) {
+                    DiskLoc next = x.btree()->tempNext();
+                    string ns = idx.indexNamespace();
+                    theDataFileMgr._deleteRecord(nsdetails(ns.c_str()), ns.c_str(), x.rec(), x);
+                    x = next;
+                    getDur().commitIfNeeded();
+                }
+                assert( idx.head.isNull() );
+                log(2) << "done rollback" << endl;
             }
-            assert( idx.head.isNull() );
-            log(2) << "done rollback" << endl;
-        }
+        )
     }
 
 }

db/client.cpp

@@ -169,11 +169,6 @@ namespace mongo {
             uassert( 13005 , "can't create db, keeps getting closed" , _db );
         }
 
-        _client->_context = this;
-        _client->_curOp->enter( this );
-        if ( doauth )
-            _auth( lockState );
-
         switch ( _client->_curOp->getOp() ) {
         case dbGetMore: // getMore's are special and should be handled else where
         case dbUpdate: // update & delete check shard version in instance.cpp, so don't check here as well
@@ -188,6 +183,11 @@ namespace mongo {
             }
         }
         }
+
+        _client->_context = this;
+        _client->_curOp->enter( this );
+        if ( doauth )
+            _auth( lockState );
     }
 
     void Client::Context::_auth( int lockState ) {

db/clientcursor.h

@@ -91,6 +91,13 @@ namespace mongo {
                     _c = 0;
                 }
             }
+            /**
+             * call this if during a yield, the cursor got deleted
+             * if so, we don't want to use the point address
+             */
+            void deleted() {
+                _c = 0;
+            }
             ~Pointer() { release(); }
             Pointer(long long cursorid) {
                 recursive_scoped_lock lock(ccmutex);

db/cloner.cpp

@@ -244,6 +244,7 @@ namespace mongo {
             string temp = ctx.db()->name + ".system.indexes";
             copy( temp.c_str() , temp.c_str() , /*isindex*/true , logForRepl , false , true , BSON( "ns" << ns ) );
         }
+        getDur().commitIfNeeded();
         return true;
     }
 

db/db.cpp

@@ -723,7 +723,11 @@ int main(int argc, char* argv[]) {
     dbExecCommand = argv[0];
 
     srand(curTimeMicros());
+#if( BOOST_VERSION >= 104500 )
+    boost::filesystem::path::default_name_check( boost::filesystem2::no_check );
+#else
     boost::filesystem::path::default_name_check( boost::filesystem::no_check );
+#endif
 
     {
         unsigned x = 0x12345678;
@@ -1099,6 +1103,11 @@ namespace mongo {
         oss << "Backtrace:" << endl;
         printStackTrace( oss );
         rawOut( oss.str() );
+
+        if( cmdLine.dur ) { 
+            ::exit(EXIT_ABRUPT);
+        }
+
         dbexit( EXIT_ABRUPT );
     }
 

db/dbcommands.cpp

@@ -406,6 +406,8 @@ namespace mongo {
                     t.appendBool( "supported" , false );
                 }
 
+                timeBuilder.appendNumber( "middle of mem" , Listener::getElapsedTimeMillis() - start );
+
                 t.appendNumber( "mapped" , MemoryMappedFile::totalMappedLength() / ( 1024 * 1024 ) );
 
                 t.done();

db/dur.cpp

@@ -211,14 +211,14 @@ namespace mongo {
             return p;
         }
 
-        void DurableImpl::commitIfNeeded() {
-#if defined(_DEBUG)
-            commitJob._nSinceCommitIfNeededCall = 0;
-#endif
+        bool DurableImpl::commitIfNeeded() {
+            DEV commitJob._nSinceCommitIfNeededCall = 0;
             if (commitJob.bytes() > UncommittedBytesLimit) { // should this also fire if CmdLine::DurAlwaysCommit?
                 stats.curr->_earlyCommits++;
                 groupCommit();
+                return true;
             }
+            return false;
         }
 
         /** Used in _DEBUG builds to check that we didn't overwrite the last intent
@@ -369,7 +369,7 @@ namespace mongo {
             // faults after remapping, so doing a little bit at a time will avoid big load spikes on
             // remapping.
             unsigned long long now = curTimeMicros64();
-            double fraction = (now-lastRemap)/20000000.0;
+            double fraction = (now-lastRemap)/2000000.0;
             lastRemap = now;
 
             rwlock lk(MongoFile::mmmutex, false);
@@ -522,7 +522,9 @@ namespace mongo {
             }
 
             // starvation on read locks could occur.  so if read lock acquisition is slow, try to get a
-            // write lock instead.  otherwise writes could use too much RAM.
+            // write lock instead.  otherwise journaling could be delayed too long (too much data will 
+            // not accumulate though, as commitIfNeeded logic will have executed in the meantime if there 
+            // has been writes)
             writelock lk;
             groupCommit();
         }
@@ -587,18 +589,11 @@ namespace mongo {
         void recover();
 
         void releasingWriteLock() {
-            try {
-#if defined(_DEBUG)
-                commitJob._nSinceCommitIfNeededCall = 0; // implicit commit if needed
-#endif
-                if (commitJob.bytes() > UncommittedBytesLimit || cmdLine.durOptions & CmdLine::DurAlwaysCommit) {
-                    stats.curr->_earlyCommits++;
-                    groupCommit();
-                }
-            }
-            catch(std::exception& e) {
-                log() << "exception in dur::releasingWriteLock causing immediate shutdown: " << e.what() << endl;
-                abort(); // based on myTerminate()
+            // implicit commitIfNeeded check on each write unlock
+            DEV commitJob._nSinceCommitIfNeededCall = 0; // implicit commit if needed
+            if( commitJob.bytes() > UncommittedBytesLimit || cmdLine.durOptions & CmdLine::DurAlwaysCommit ) {
+                stats.curr->_earlyCommits++;
+                groupCommit();
             }
         }
 

db/dur.h

@@ -96,8 +96,9 @@ namespace mongo {
                 (like creating an index or update with $atomic) can call this
                 whenever the db is in a sane state and it will prevent commits
                 from growing too large.
+                @return true if commited
             */
-            virtual void commitIfNeeded() = 0;
+            virtual bool commitIfNeeded() = 0;
 
             /** Declare write intent for a DiskLoc.  @see DiskLoc::writing() */
             inline DiskLoc& writingDiskLoc(DiskLoc& d) { return *((DiskLoc*) writingPtr(&d, sizeof(d))); }
@@ -172,7 +173,7 @@ namespace mongo {
             void createdFile(string filename, unsigned long long len) { }
             bool awaitCommit() { return false; }
             bool commitNow() { return false; }
-            void commitIfNeeded() { }
+            bool commitIfNeeded() { return false; }
             void setNoJournal(void *dst, void *src, unsigned len);
             void syncDataAndTruncateJournal() {}
         };
@@ -185,7 +186,7 @@ namespace mongo {
             void createdFile(string filename, unsigned long long len);
             bool awaitCommit();
             bool commitNow();
-            void commitIfNeeded();
+            bool commitIfNeeded();
             void setNoJournal(void *dst, void *src, unsigned len);
             void syncDataAndTruncateJournal();
         };

db/dur_journal.cpp

@@ -262,6 +262,7 @@ namespace mongo {
                         log() << "warning caught exception in preallocateFiles, continuing" << endl;
                     }
             }
+            j.open();
         }
 
         void removeOldJournalFile(path p) { 
@@ -515,7 +516,7 @@ namespace mongo {
 
             scoped_lock lk(_curLogFileMutex);
 
-            if ( inShutdown() )
+            if ( inShutdown() || !_curLogFile )
                 return;
 
             j.updateLSNFile();

db/dur_journalimpl.h

@@ -63,10 +63,10 @@ namespace mongo {
                     _open();
             }
 
-        private:
             /** open a journal file to journal operations to. */
             void open();
 
+        private:
             void _open();
             void closeCurrentJournalFile();
             void removeUnneededJournalFiles();

db/dur_recover.cpp

@@ -73,16 +73,17 @@ namespace mongo {
                     if( m.count(u) ) {
                         uasserted(13531, str::stream() << "unexpected files in journal directory " << dir.string() << " : " << fileName);
                     }
-                    if( !m.empty() && !m.count(u-1) ) {
-                        uasserted(13532,
-                                  str::stream() << "unexpected file in journal directory " << dir.string()
-                                  << " : " << fileName << " : can't find its preceeding file");
-                    }
                     m.insert( pair<unsigned,path>(u,filepath) );
                 }
             }
-            for( map<unsigned,path>::iterator i = m.begin(); i != m.end(); ++i )
+            for( map<unsigned,path>::iterator i = m.begin(); i != m.end(); ++i ) {
+                if( i != m.begin() && m.count(i->first - 1) == 0 ) {
+                    uasserted(13532,
+                    str::stream() << "unexpected file in journal directory " << dir.string()
+                      << " : " << filesystem::path(i->second).leaf() << " : can't find its preceeding file");
+                }
                 files.push_back(i->second);
+            }
         }
 
         /** read through the memory mapped data of a journal file (journal/j._<n> file)
@@ -186,8 +187,10 @@ namespace mongo {
         }
 
         RecoveryJob::~RecoveryJob() {
-            if( !_mmfs.empty() )
-                close();
+            DESTRUCTOR_GUARD(
+                if( !_mmfs.empty() )
+                    close();
+            )
         }
 
         void RecoveryJob::close() {
@@ -445,8 +448,8 @@ namespace mongo {
             }
         } brunittest;
 
-
-        RecoveryJob RecoveryJob::_instance;
+        // can't free at termination because order of destruction of global vars is arbitrary
+        RecoveryJob &RecoveryJob::_instance = *(new RecoveryJob());
 
     } // namespace dur
 

db/dur_recover.h

@@ -15,7 +15,7 @@ namespace mongo {
          */
         class RecoveryJob : boost::noncopyable {
         public:
-            RecoveryJob() :_lastDataSyncedFromLastRun(0), _mx("recovery") { _lastSeqMentionedInConsoleLog = 1; }
+            RecoveryJob() :_lastDataSyncedFromLastRun(0), _mx("recovery"), _recovering(false) { _lastSeqMentionedInConsoleLog = 1; }
             void go(vector<path>& files);
             ~RecoveryJob();
             void processSection(const void *, unsigned len);
@@ -39,7 +39,7 @@ namespace mongo {
 
             bool _recovering; // are we in recovery or WRITETODATAFILES
 
-            static RecoveryJob _instance;
+            static RecoveryJob &_instance;
         };
     }
 }

db/mongommf.cpp

@@ -342,11 +342,7 @@ namespace mongo {
 
     bool MongoMMF::create(string fname, unsigned long long& len, bool sequentialHint) {
         setPath(fname);
-        bool preExisting = MemoryMappedFile::exists(fname.c_str());
         _view_write = map(fname.c_str(), len, sequentialHint ? SEQUENTIAL : 0);
-        if( cmdLine.dur && _view_write && !preExisting ) {
-            getDur().createdFile(fname, len);
-        }
         return finishOpening();
     }
 

db/namespace.cpp

@@ -182,6 +182,7 @@ namespace mongo {
             maybeMkdir();
             unsigned long long l = lenForNewNsFiles;
             if( f.create(pathString, l, true) ) {
+                getDur().createdFile(pathString, l); // always a new file
                 len = l;
                 assert( len == lenForNewNsFiles );
                 p = f.getView();
@@ -194,6 +195,7 @@ namespace mongo {
             dbexit( EXIT_FS );
         }
 
+
         assert( len <= 0x7fffffff );
         ht = new HashTable<Namespace,NamespaceDetails>(p, (int) len, "namespace index");
         if( checkNsFilesOnLoad )

db/pdfile.cpp

@@ -252,6 +252,10 @@ namespace mongo {
             while ( size > 0 ) {
                 int max = MongoDataFile::maxSize() - DataFileHeader::HeaderSize;
                 int desiredExtentSize = (int) (size > max ? max : size);
+                if ( desiredExtentSize < Extent::minSize() ) {
+                    desiredExtentSize = Extent::minSize();
+                }
+                desiredExtentSize &= 0xffffff00;
                 Extent *e = database->allocExtent( ns, desiredExtentSize, newCapped );
                 size -= e->length;
             }
@@ -403,7 +407,7 @@ namespace mongo {
             uassert( 10084 , "can't map file memory - mongo requires 64 bit build for larger datasets", _mb != 0);
         else
             uassert( 10085 , "can't map file memory", _mb != 0);
-        header()->init(fileNo, size);
+        header()->init(fileNo, size, filename);
     }
 
     void MongoDataFile::flush( bool sync ) {
@@ -435,11 +439,11 @@ namespace mongo {
 
     Extent* MongoDataFile::createExtent(const char *ns, int approxSize, bool newCapped, int loops) {
         massert( 10357 ,  "shutdown in progress", ! inShutdown() );
-        massert( 10358 ,  "bad new extent size", approxSize >= 0 && approxSize <= Extent::maxSize() );
+        massert( 10358 ,  "bad new extent size", approxSize >= Extent::minSize() && approxSize <= Extent::maxSize() );
         massert( 10359 ,  "header==0 on new extent: 32 bit mmap space exceeded?", header() ); // null if file open failed
         int ExtentSize = approxSize <= header()->unusedLength ? approxSize : header()->unusedLength;
         DiskLoc loc;
-        if ( ExtentSize <= 0 ) {
+        if ( ExtentSize < Extent::minSize() ) {
             /* not there could be a lot of looping here is db just started and
                no files are open yet.  we might want to do something about that. */
             if ( loops > 8 ) {
@@ -1187,7 +1191,9 @@ namespace mongo {
             op->setMessage( "index: (3/3) btree-middle" );
             log(t.seconds() > 10 ? 0 : 1 ) << "\t done building bottom layer, going to commit" << endl;
             btBuilder.commit();
-            wassert( btBuilder.getn() == nkeys || dropDups );
+            if ( btBuilder.getn() != nkeys && ! dropDups ) {
+                warning() << "not all entries were added to the index, probably some keys were too large" << endl;
+            }
         }
 
         log(1) << "\t fastBuildIndex dupsToDrop:" << dupsToDrop.size() << endl;

db/pdfile.h

@@ -272,6 +272,7 @@ namespace mongo {
         Extent* getPrevExtent() { return xprev.isNull() ? 0 : DataFileMgr::getExtent(xprev); }
 
         static int maxSize();
+        static int minSize() { return 0x100; }
         /**
          * @param len lengt of record we need
          * @param lastRecord size of last extent which is a factor in next extent size
@@ -325,11 +326,12 @@ namespace mongo {
 
         bool uninitialized() const { return version == 0; }
 
-        void init(int fileno, int filelength) {
+        void init(int fileno, int filelength, const char* filename) {
             if ( uninitialized() ) {
                 if( !(filelength > 32768 ) ) { 
                     massert(13640, str::stream() << "DataFileHeader looks corrupt at file open filelength:" << filelength << " fileno:" << fileno, false);
                 }
+                getDur().createdFile(filename, filelength);
                 assert( HeaderSize == 8192 );
                 DataFileHeader *h = getDur().writing(this);
                 h->fileLength = filelength;

db/query.cpp

@@ -380,8 +380,16 @@ namespace mongo {
                     }
                 }
                 c->advance();
-            }
 
+                if ( ! cc->yieldSometimes() ) {
+                    ClientCursor::erase(cursorid);
+                    cursorid = 0;
+                    cc = 0;
+                    p.deleted();
+                    break;
+                }
+            }
+            
             if ( cc ) {
                 cc->updateLocation();
                 cc->mayUpgradeStorage();

db/repl.cpp

@@ -1426,6 +1426,8 @@ namespace mongo {
                         break;
                     }
                     op = oplogReader.next();
+
+                    getDur().commitIfNeeded();
                 }
             }
         }

db/repl/rs_sync.cpp

@@ -143,6 +143,8 @@ namespace mongo {
                     // simple progress metering
                     log() << "replSet initialSyncOplogApplication " << n << rsLog;
                 }
+                
+                getDur().commitIfNeeded();
             }
             catch (DBException& e) {
                 if( e.getCode() == 11000 || e.getCode() == 11001 ) {

db/stats/top.cpp

@@ -44,6 +44,9 @@ namespace mongo {
     }
 
     void Top::record( const string& ns , int op , int lockType , long long micros , bool command ) {
+        if ( ns[0] == '?' )
+            return;
+
         //cout << "record: " << ns << "\t" << op << "\t" << command << endl;
         scoped_lock lk(_lock);
 

debian/changelog

@@ -1,3 +1,9 @@
+mongodb (1.8.0-rc1) unstable; urgency=low
+
+  * see http://jira.mongodb.org/browse/SERVER/fixforversion/10246
+
+ -- Richard Kreuter <richard@10gen.com>  Tue, 1 Mar 2011 16:56:28 -0500
+
 mongodb (1.8.0-rc0) unstable; urgency=low
 
   * see http://jira.mongodb.org/browse/SERVER/fixforversion/10245

doxygenConfig

@@ -3,7 +3,7 @@
 #---------------------------------------------------------------------------
 DOXYFILE_ENCODING      = UTF-8
 PROJECT_NAME           = MongoDB
-PROJECT_NUMBER         = 1.8.0-rc0
+PROJECT_NUMBER         = 1.8.0-rc2
 OUTPUT_DIRECTORY       = docs/doxygen
 CREATE_SUBDIRS         = NO
 OUTPUT_LANGUAGE        = English

jstests/check_shard_index.js

@@ -32,9 +32,14 @@ f.ensureIndex( { x: 1 , y: 1 } );
 assert.eq( 0 , f.count() , "2. initial count should be zero" );
 
 f.save( { x: 1 , y : 1 } );
-f.save( { y: 2 } );
-assert.eq( 2 , f.count() , "2. count after initial insert should be 2" );
+f.save( { x: null , y : 1 } );
+
 res = db.runCommand( { checkShardingIndex: "test.jstests_shardingindex" , keyPattern: {x:1, y:1} , force: true });
-assert.eq( false , res.ok , "2a" );
+assert.eq( true , res.ok , "2a " + tojson(res) );
+
+f.save( { y: 2 } );
+assert.eq( 3 , f.count() , "2. count after initial insert should be 3" );
+res = db.runCommand( { checkShardingIndex: "test.jstests_shardingindex" , keyPattern: {x:1, y:1} , force: true });
+assert.eq( false , res.ok , "2b " + tojson(res) );
 
 print("PASSED");

jstests/index11.js

@@ -5,12 +5,9 @@ coll.drop();
 
 coll.ensureIndex({"k": 1, "v": 1});
 coll.insert({k: "x", v: "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"});
-/* TODO inconsistent behavior w/ too large to index - this fails but the next assert works.
- * SERVER-1716
- * assert.eq(1, coll.find({"k": "x"}).count());
- */
+assert.eq(0, coll.find({"k": "x"}).count()); // SERVER-1716
 
 coll.dropIndexes();
 coll.ensureIndex({"k": 1, "v": 1});
 
-assert.eq(1, coll.find({"k": "x"}).count());
+assert.eq(0, coll.find({"k": "x"}).count());

jstests/sharding/features1.js

@@ -175,7 +175,7 @@ assert( ! s.admin.runCommand( { shardcollection : "test.foo9" , key : { a : 1 }
 // --- listDatabases ---
 
 r = db.getMongo().getDBs()
-assert.eq( 4 , r.databases.length , "listDatabases 1 : " + tojson( r ) )
+assert.eq( 3 , r.databases.length , "listDatabases 1 : " + tojson( r ) )
 assert.lt( 10000 , r.totalSize , "listDatabases 2 : " + tojson( r ) );
 
 s.stop()

jstests/slowNightly/newcollection2.js

@@ -0,0 +1,11 @@
+// Alocate collection forcing just a small size remainder in 2nd extent
+
+port = allocatePorts( 1 )[ 0 ]
+var baseName = "jstests_disk_newcollection2";
+var m = startMongod( "--noprealloc", "--smallfiles", "--port", port, "--dbpath", "/data/db/" + baseName );
+db = m.getDB( "test" );
+
+db.createCollection( baseName, {size:0x1FFC0000-0x10-8192} );
+var v = db[ baseName ].validate();
+printjson( v );
+assert( v.valid );

jstests/slowNightly/sharding_rs1.js

@@ -43,6 +43,15 @@ function diff(){
 assert.lt( 20 , diff() , "big differential here" );
 print( diff() )
 
+{
+    // quick test for SERVER-2686
+    var mydbs = db.getMongo().getDBs().databases;
+    for ( var i=0; i<mydbs.length; i++ ) {
+        assert( mydbs[i].name != "local" , "mongos listDatabases can't return local" );
+    }
+}
+
+
 assert.soon( function(){
     var d = diff();
     return d < 5;

rpm/mongo.spec

@@ -1,5 +1,5 @@
 Name: mongo
-Version: 1.8.0-rc0
+Version: 1.8.0-rc1
 Release: mongodb_1%{?dist}
 Summary: mongo client shell and tools
 License: AGPL 3.0

s/balance.cpp

@@ -275,6 +275,15 @@ namespace mongo {
         while ( ! inShutdown() ) {
 
             try {
+                
+                // first make sure we should even be running
+                if ( ! grid.shouldBalance() ) {
+                    log(1) << "skipping balancing round because balancing is disabled" << endl;
+                    sleepsecs( 30 );
+                    continue;
+                }
+                
+
                 ScopedDbConnection conn( config );
 
                 _ping( conn.conn() );
@@ -294,14 +303,6 @@ namespace mongo {
                     continue;
                 }
 
-                if ( ! grid.shouldBalance() ) {
-                    log(1) << "skipping balancing round because balancing is disabled" << endl;;
-                    conn.done();
-
-                    sleepsecs( 30 );
-                    continue;
-                }
-
                 log(1) << "*** start balancing round" << endl;
 
                 vector<CandidateChunkPtr> candidateChunks;

s/balancer_policy.cpp

@@ -155,7 +155,7 @@ namespace mongo {
 
     bool BalancerPolicy::isDraining( BSONObj limits ) {
         BSONElement draining = limits[ ShardFields::draining.name() ];
-        if ( draining.eoo() || ! draining.Bool() ) {
+        if ( draining.eoo() || ! draining.trueValue() ) {
             return false;
         }
 
@@ -164,7 +164,7 @@ namespace mongo {
 
     bool BalancerPolicy::hasOpsQueued( BSONObj limits ) {
         BSONElement opsQueued = limits[ LimitsFields::hasOpsQueued.name() ];
-        if ( opsQueued.eoo() || ! opsQueued.Bool() ) {
+        if ( opsQueued.eoo() || ! opsQueued.trueValue() ) {
             return false;
         }
         return true;

s/commands_admin.cpp

@@ -534,7 +534,7 @@ namespace mongo {
                     result.append( "cause" , res );
                     return false;
                 }
-
+                config->getChunkManager( ns , true );
                 return true;
             }
         } splitCollectionCmd;
@@ -959,6 +959,13 @@ namespace mongo {
             BSONArrayBuilder bb( result.subarrayStart( "databases" ) );
             for ( map<string,long long>::iterator i=sizes.begin(); i!=sizes.end(); ++i ) {
                 string name = i->first;
+
+                if ( name == "local" ) {
+                    // we don't return local
+                    // since all shards have their own independant local
+                    continue;
+                }
+
                 long long size = i->second;
                 totalSize += size;
 

s/d_migrate.cpp

@@ -1162,7 +1162,7 @@ namespace mongo {
                 string system_indexes = cc().database()->name + ".system.indexes";
                 for ( unsigned i=0; i<all.size(); i++ ) {
                     BSONObj idx = all[i];
-                    theDataFileMgr.insert( system_indexes.c_str() , idx.objdata() , idx.objsize() );
+                    theDataFileMgr.insertAndLog( system_indexes.c_str() , idx );
                 }
 
                 timing.done(1);

s/d_split.cpp

@@ -178,18 +178,37 @@ namespace mongo {
             }
 
             // for now, the only check is that all shard keys are filled
+            // null is ok, 
             // TODO if $exist for nulls were picking the index, it could be used instead efficiently
             while ( cc->ok() ) {
                 BSONObj currKey = c->currKey();
-                BSONForEach(key, currKey) {
-                    if ( key.type() == jstNULL ) {
-                        ostringstream os;
-                        os << "found null value in key " << bc->prettyKey( currKey );
-                        log() << "checkShardingIndex for '" << ns << "' failed: " << os.str() << endl;
+                
+                BSONObjIterator i( currKey );
+                int n = 0;
+                while ( i.more() ) {
+                    BSONElement key = i.next();
+                    n++;
 
-                        errmsg = os.str();
-                        return false;
-                    }
+                    if ( key.type() && key.type() != jstNULL )
+                        continue;
+                    
+                    BSONObj obj = c->current();
+                    BSONObjIterator j( keyPattern );
+                    BSONElement real;
+                    for ( int x=0; x<n; x++ )
+                        real = j.next();
+                    
+                    real = obj.getFieldDotted( real.fieldName() );
+
+                    if ( real.type() )
+                        continue;
+                    
+                    ostringstream os;
+                    os << "found null value in key " << bc->prettyKey( currKey ) << " for doc: " << real["_id"];
+                    log() << "checkShardingIndex for '" << ns << "' failed: " << os.str() << endl;
+                    
+                    errmsg = os.str();
+                    return false;
                 }
                 cc->advance();
             }
@@ -353,48 +372,61 @@ namespace mongo {
                 // at the end. If a key appears more times than entries allowed on a chunk, we issue a warning and
                 // split on the following key.
                 set<BSONObj> tooFrequentKeys;
-                splitKeys.push_back( c->currKey() );
-                while ( cc->ok() ) {
-                    currCount++;
-                    BSONObj currKey = c->currKey();
-                    
-                    DEV assert( currKey.woCompare( max ) <= 0 );
-                    
-                    if ( currCount > keyCount ) {
-                        // Do not use this split key if it is the same used in the previous split point.
-                        if ( currKey.woCompare( splitKeys.back() ) == 0 ) {
-                            tooFrequentKeys.insert( currKey.getOwned() );
+                splitKeys.push_back( c->currKey().getOwned() );
+                while ( 1 ) {
+                    while ( cc->ok() ) {
+                        currCount++;
+                        BSONObj currKey = c->currKey();
+                        
+                        DEV assert( currKey.woCompare( max ) <= 0 );
+                        
+                        if ( currCount > keyCount ) {
+                            // Do not use this split key if it is the same used in the previous split point.
+                            if ( currKey.woCompare( splitKeys.back() ) == 0 ) {
+                                tooFrequentKeys.insert( currKey.getOwned() );
+                                
+                            }
+                            else {
+                                splitKeys.push_back( currKey.getOwned() );
+                                currCount = 0;
+                                numChunks++;
+                                
+                                LOG(4) << "picked a split key: " << bc->prettyKey( currKey ) << endl;
+                            }
                             
                         }
-                        else {
-                            splitKeys.push_back( currKey.getOwned() );
-                            currCount = 0;
-                            numChunks++;
-                            
-                            LOG(4) << "picked a split key: " << bc->prettyKey( currKey ) << endl;
+                        
+                        cc->advance();
+                        
+                        // Stop if we have enough split points.
+                        if ( maxSplitPoints && ( numChunks >= maxSplitPoints ) ) {
+                            log() << "max number of requested split points reached (" << numChunks
+                                  << ") before the end of chunk " << ns << " " << min << " -->> " << max
+                                  << endl;
+                            break;
                         }
                         
-                    }
-
-                    cc->advance();
-                    
-                    // Stop if we have enough split points.
-                    if ( maxSplitPoints && ( numChunks >= maxSplitPoints ) ) {
-                        log() << "max number of requested split points reached (" << numChunks
-                              << ") before the end of chunk " << ns << " " << min << " -->> " << max
-                              << endl;
-                        break;
+                        if ( ! cc->yieldSometimes() ) {
+                            // we were near and and got pushed to the end
+                            // i think returning the splits we've already found is fine
+                            
+                            // don't use the btree cursor pointer to acces keys beyond this point but ok
+                            // to use it for format the keys we've got already
+                            
+                            break;
+                        }
                     }
                     
-                    if ( ! cc->yieldSometimes() ) {
-                        // we were near and and got pushed to the end
-                        // i think returning the splits we've already found is fine
-                        
-                        // don't use the btree cursor pointer to acces keys beyond this point but ok
-                        // to use it for format the keys we've got already
-                        
+                    if ( splitKeys.size() > 1 || ! force )
                         break;
-                    }
+                    
+                    force = false;
+                    keyCount = currCount / 2;
+                    currCount = 0;
+                    log() << "splitVector doing another cycle because of force, keyCount now: " << keyCount << endl;
+                    
+                    c.reset( new BtreeCursor( d , d->idxNo(*idx) , *idx , min , max , false , 1 ) );
+                    cc.reset( new ClientCursor( QueryOption_NoCursorTimeout , c , ns ) );
                 }
                 
                 //

s/d_state.cpp

@@ -60,6 +60,7 @@ namespace mongo {
     }
 
     void ShardingState::gotShardName( const string& name ) {
+        scoped_lock lk(_mutex);
         if ( _shardName.size() == 0 ) {
             // TODO SERVER-2299 verify the name is sound w.r.t IPs
             _shardName = name;
@@ -78,7 +79,7 @@ namespace mongo {
     }
 
     void ShardingState::gotShardHost( string host ) {
-
+        scoped_lock lk(_mutex);
         size_t slash = host.find( '/' );
         if ( slash != string::npos )
             host = host.substr( 0 , slash );
@@ -386,79 +387,128 @@ namespace mongo {
             help << " example: { setShardVersion : 'alleyinsider.foo' , version : 1 , configdb : '' } ";
         }
 
-        virtual LockType locktype() const { return WRITE; } // TODO: figure out how to make this not need to lock
+        virtual LockType locktype() const { return NONE; }
+        
+        bool checkConfigOrInit( const string& configdb , bool authoritative , string& errmsg , BSONObjBuilder& result , bool locked=false ) const {
+            if ( configdb.size() == 0 ) {
+                errmsg = "no configdb";
+                return false;
+            }
+            
+            if ( shardingState.enabled() ) {
+                if ( configdb == shardingState.getConfigServer() ) 
+                    return true;
+                
+                result.append( "configdb" , BSON( "stored" << shardingState.getConfigServer() << 
+                                                  "given" << configdb ) );
+                errmsg = "specified a different configdb!";
+                return false;
+            }
+            
+            if ( ! authoritative ) {
+                result.appendBool( "need_authoritative" , true );
+                errmsg = "first setShardVersion";
+                return false;
+            }
+            
+            if ( locked ) {
+                shardingState.enable( configdb );
+                configServer.init( configdb );
+                return true;
+            }
+
+            dblock lk;
+            return checkConfigOrInit( configdb , authoritative , errmsg , result , true );
+        }
+        
+        bool checkMongosID( ShardedConnectionInfo* info, const BSONElement& id, string errmsg ) {
+            if ( id.type() != jstOID ) {
+                // TODO: fix this
+                //errmsg = "need serverID to be an OID";
+                //return 0;
+                return true;
+            }
+            
+            OID clientId = id.__oid();
+            if ( ! info->hasID() ) {
+                info->setID( clientId );
+                return true;
+            }
+            
+            if ( clientId != info->getID() ) {
+                errmsg = "server id has changed!";
+                return false;
+            }
+
+            return true;
+        }
 
         bool run(const string& , BSONObj& cmdObj, string& errmsg, BSONObjBuilder& result, bool) {
 
+            // Steps
+            // 1. check basic config
+            // 2. extract params from command
+            // 3. fast check
+            // 4. slow check (LOCKS)
+            
+            // step 1
+
             lastError.disableForCommand();
             ShardedConnectionInfo* info = ShardedConnectionInfo::get( true );
 
             bool authoritative = cmdObj.getBoolField( "authoritative" );
+            
+            // check config server is ok or enable sharding
+            if ( ! checkConfigOrInit( cmdObj["configdb"].valuestrsafe() , authoritative , errmsg , result ) )
+                return false;
 
-            string configdb = cmdObj["configdb"].valuestrsafe();
-            {
-                // configdb checking
-                if ( configdb.size() == 0 ) {
-                    errmsg = "no configdb";
-                    return false;
-                }
-
-                if ( shardingState.enabled() ) {
-                    if ( configdb != shardingState.getConfigServer() ) {
-                        errmsg = "specified a different configdb!";
-                        return false;
-                    }
-                }
-                else {
-                    if ( ! authoritative ) {
-                        result.appendBool( "need_authoritative" , true );
-                        errmsg = "first setShardVersion";
-                        return false;
-                    }
-                    shardingState.enable( configdb );
-                    configServer.init( configdb );
-                }
-            }
-
+            // check shard name/hosts are correct
             if ( cmdObj["shard"].type() == String ) {
                 shardingState.gotShardName( cmdObj["shard"].String() );
                 shardingState.gotShardHost( cmdObj["shardHost"].String() );
             }
-
-            {
-                // setting up ids
-                if ( cmdObj["serverID"].type() != jstOID ) {
-                    // TODO: fix this
-                    //errmsg = "need serverID to be an OID";
-                    //return 0;
-                }
-                else {
-                    OID clientId = cmdObj["serverID"].__oid();
-                    if ( ! info->hasID() ) {
-                        info->setID( clientId );
-                    }
-                    else if ( clientId != info->getID() ) {
-                        errmsg = "server id has changed!";
-                        return 0;
-                    }
-                }
-            }
-
-            unsigned long long version = extractVersion( cmdObj["version"] , errmsg );
-
-            if ( errmsg.size() ) {
+            
+            // make sure we have the mongos id for writebacks
+            if ( ! checkMongosID( info , cmdObj["serverID"] , errmsg ) )
                 return false;
-            }
 
+            // step 2
+            
             string ns = cmdObj["setShardVersion"].valuestrsafe();
             if ( ns.size() == 0 ) {
-                errmsg = "need to speciy fully namespace";
+                errmsg = "need to speciy namespace";
                 return false;
             }
 
+            const ConfigVersion version = extractVersion( cmdObj["version"] , errmsg );
+            if ( errmsg.size() )
+                return false;
+            
+            // step 3
+
             const ConfigVersion oldVersion = info->getVersion(ns);
             const ConfigVersion globalVersion = shardingState.getVersion(ns);
 
+            result.appendTimestamp( "oldVersion" , oldVersion );
+            
+            if ( globalVersion > 0 && version > 0 ) {
+                // this means there is no reset going on an either side
+                // so its safe to make some assuptions
+
+                if ( version == globalVersion ) {
+                    // mongos and mongod agree!
+                    if ( oldVersion != version ) {
+                        assert( oldVersion < globalVersion );
+                        info->setVersion( ns , version );
+                    }
+                    return true;
+                }
+                
+            }
+
+            // step 4
+            dblock setShardVersionLock; // TODO: can we get rid of this??
+            
             if ( oldVersion > 0 && globalVersion == 0 ) {
                 // this had been reset
                 info->setVersion( ns , 0 );
@@ -475,7 +525,6 @@ namespace mongo {
                     result.appendBool( "need_authoritative" , true );
                     result.append( "ns" , ns );
                     result.appendTimestamp( "globalVersion" , globalVersion );
-                    result.appendTimestamp( "oldVersion" , oldVersion );
                     errmsg = "dropping needs to be authoritative";
                     return false;
                 }
@@ -491,7 +540,6 @@ namespace mongo {
             if ( version < oldVersion ) {
                 errmsg = "you already have a newer version of collection '" + ns + "'";
                 result.append( "ns" , ns );
-                result.appendTimestamp( "oldVersion" , oldVersion );
                 result.appendTimestamp( "newVersion" , version );
                 result.appendTimestamp( "globalVersion" , globalVersion );
                 return false;
@@ -510,7 +558,7 @@ namespace mongo {
                 return false;
             }
 
-            if ( globalVersion == 0 && ! cmdObj.getBoolField( "authoritative" ) ) {
+            if ( globalVersion == 0 && ! authoritative ) {
                 // need authoritative for first look
                 result.append( "ns" , ns );
                 result.appendBool( "need_authoritative" , true );
@@ -518,6 +566,7 @@ namespace mongo {
                 return false;
             }
 
+            Timer relockTime;
             {
                 dbtemprelease unlock;
 
@@ -530,11 +579,11 @@ namespace mongo {
                     return false;
                 }
             }
-
+            if ( relockTime.millis() >= ( cmdLine.slowMS - 10 ) ) {
+                log() << "setShardVersion - relocking slow: " << relockTime.millis() << endl;
+            }
+            
             info->setVersion( ns , version );
-            result.appendTimestamp( "oldVersion" , oldVersion );
-            result.append( "ok" , 1 );
-
             return true;
         }
 
@@ -635,10 +684,10 @@ namespace mongo {
             return false;
         }
 
-        if ( isWriteOp && version.majorVersion() == clientVersion.majorVersion() ) {
+        if ( version.majorVersion() == clientVersion.majorVersion() ) {
             // this means there was just a split
             // since on a split w/o a migrate this server is ok
-            // going to accept write
+            // going to accept 
             return true;
         }
 

shell/db.js

@@ -769,8 +769,8 @@ DB.prototype.listCommands = function(){
     }
 }
 
-DB.prototype.printShardingStatus = function(){
-    printShardingStatus( this.getSiblingDB( "config" ) );
+DB.prototype.printShardingStatus = function( verbose ){
+    printShardingStatus( this.getSiblingDB( "config" ) , verbose );
 }
 
 DB.autocomplete = function(obj){

shell/servers.js

@@ -478,7 +478,7 @@ printShardingStatus = function( configDB , verbose ){
                                 output( "\t\t\t\t" + z.shard + "\t" + z.nChunks );
                             } )
                             
-                            if ( totalChunks < 1000 || verbose ){
+                            if ( totalChunks < 20 || verbose ){
                                 configDB.chunks.find( { "ns" : coll._id } ).sort( { min : 1 } ).forEach( 
                                     function(chunk){
                                         output( "\t\t\t" + tojson( chunk.min ) + " -->> " + tojson( chunk.max ) + 

tools/dump.cpp

@@ -69,13 +69,13 @@ public:
         Writer writer(out, m);
 
         // use low-latency "exhaust" mode if going over the network
-        if (typeid(connBase) == typeid(DBClientConnection&)) {
+        if (!_usingMongos && typeid(connBase) == typeid(DBClientConnection&)) {
             DBClientConnection& conn = static_cast<DBClientConnection&>(connBase);
             boost::function<void(const BSONObj&)> castedWriter(writer); // needed for overload resolution
             conn.query( castedWriter, coll.c_str() , q , NULL, queryOptions | QueryOption_Exhaust);
         }
         else {
-            //This branch should only be taken with DBDirectClient which doesn't support exhaust mode
+            //This branch should only be taken with DBDirectClient or mongos which doesn't support exhaust mode
             scoped_ptr<DBClientCursor> cursor(connBase.query( coll.c_str() , q , 0 , 0 , 0 , queryOptions ));
             while ( cursor->more() ) {
                 writer(cursor->next());
@@ -303,8 +303,6 @@ public:
             opLogStart = op["ts"]._numberLong();
         }
 
-
-
         // check if we're outputting to stdout
         string out = getParam("out");
         if ( out == "-" ) {
@@ -318,6 +316,13 @@ public:
             }
         }
 
+        {
+            // TODO: when mongos supports QueryOption_Exaust add a version check (SERVER-2628)
+            BSONObj isdbgrid;
+            conn("true").simpleCommand("admin", &isdbgrid, "isdbgrid");
+            _usingMongos = isdbgrid["isdbgrid"].trueValue();
+        }
+
         path root( out );
         string db = _db;
 
@@ -358,6 +363,7 @@ public:
         return 0;
     }
 
+    bool _usingMongos;
     BSONObj _query;
 };
 

tools/tool.cpp

@@ -101,7 +101,11 @@ namespace mongo {
         // we want durability to be disabled.
         cmdLine.dur = false;
 
-        boost::filesystem::path::default_name_check( boost::filesystem::no_check );
+#if( BOOST_VERSION >= 104500 )
+    boost::filesystem::path::default_name_check( boost::filesystem2::no_check );
+#else
+    boost::filesystem::path::default_name_check( boost::filesystem::no_check );
+#endif
 
         _name = argv[0];
 

util/alignedbuilder.cpp

@@ -48,7 +48,7 @@ namespace mongo {
         assert( a );
         while( 1 ) {
             a *= 2;
-            wassert( a <= 128*1024*1024 );
+            wassert( a <= 256*1024*1024 );
             assert( a <= 512*1024*1024 );
             if( _len < a )
                 break;

util/logfile.cpp

@@ -136,7 +136,7 @@ namespace mongo {
         }
         ssize_t written = write(_fd, buf, len);
         if( written != (ssize_t) len ) {
-            log() << "write fails written:" << written << " len:" << len << " buf:" << buf << " errno:" << errno << endl;
+            log() << "write fails written:" << written << " len:" << len << " buf:" << buf << ' ' << errnoWithDescription() << endl;
             uasserted(13515, str::stream() << "error appending to file " << _fd  << ' ' << errnoWithDescription());
         }
 

util/mmap_posix.cpp

@@ -51,6 +51,10 @@ namespace mongo {
 
 #ifndef O_NOATIME
 #define O_NOATIME (0)
+#endif
+
+#ifndef MAP_NORESERVE
+#define MAP_NORESERVE (0)
 #endif
 
     void* MemoryMappedFile::map(const char *filename, unsigned long long &length, int options) {
@@ -118,7 +122,7 @@ namespace mongo {
     }
 
     void* MemoryMappedFile::createPrivateMap() {
-        void * x = mmap( /*start*/0 , len , PROT_READ|PROT_WRITE , MAP_PRIVATE , fd , 0 );
+        void * x = mmap( /*start*/0 , len , PROT_READ|PROT_WRITE , MAP_PRIVATE|MAP_NORESERVE , fd , 0 );
         if( x == MAP_FAILED ) {
             if ( errno == ENOMEM ) {
                 if( sizeof(void*) == 4 ) {
@@ -140,7 +144,7 @@ namespace mongo {
 
     void* MemoryMappedFile::remapPrivateView(void *oldPrivateAddr) {
         // don't unmap, just mmap over the old region
-        void * x = mmap( oldPrivateAddr, len , PROT_READ|PROT_WRITE , MAP_PRIVATE|MAP_FIXED , fd , 0 );
+        void * x = mmap( oldPrivateAddr, len , PROT_READ|PROT_WRITE , MAP_PRIVATE|MAP_NORESERVE|MAP_FIXED , fd , 0 );
         if( x == MAP_FAILED ) {
             int err = errno;
             error()  << "13601 Couldn't remap private view: " << errnoWithDescription(err) << endl;

util/version.cpp

@@ -26,7 +26,7 @@
 
 namespace mongo {
 
-    const char versionString[] = "1.8.0-rc0";
+    const char versionString[] = "1.8.0-rc2";
 
     string mongodVersion() {
         stringstream ss;