Commit Graph

107 Commits

Author SHA1 Message Date
Eliot Horowitz
3d6f23651f fix migrate with oversize docs SERVER-2744 180 2011-03-11 16:51:14 -05:00
Eliot Horowitz
40adfd5ebd make sure that index creation is logged on secondaries during migrate SERVER-2665 2011-03-01 23:58:58 -05:00
greg
737a20b8fa Fixes for distributed lock handling time skew.
In general, the current distributed lock is susceptible to time skew in config servers and mongos clients.  This update
uses an algorithm that checks for time skew, as well as providing some more robust tests to make sure a distributed
lock cannot be forced prematurely.  Some documentation on the website, currently hidden from non-10gen, gives more
info.
2011-03-01 14:15:26 -05:00
Eliot Horowitz
f7c8abd417 fix bug in background cursor cleanup that was counting threads wrong preventing migrations back to a shard SERVER-2576 2011-02-17 11:04:51 -05:00
Eliot Horowitz
a4153eb6ed If migrate fails due to oversized chunk, try to split that chunk SERVER-2519 2011-02-16 15:51:25 -05:00
Eliot Horowitz
cfd8af10a2 make sure to init variable 2011-02-13 10:09:10 -05:00
Eliot Horowitz
44f73de7e0 better migrate commit error handling 2011-02-09 00:56:21 -05:00
Eliot Horowitz
58efac445a fix race condition in MigrateFromStatus done() 2011-02-04 23:51:17 -05:00
Eliot Horowitz
84e7e2ce17 don't allow a shard to be the TO of a migrate while its still cleaning up old data SERVER-1842 2011-01-21 19:10:51 -05:00
Alberto Lerner
8a37c45ed7 fix: commitNow requires read lock 2011-01-12 08:49:45 -05:00
Alberto Lerner
91647267b3 Revert "fix: if durability is on, mods would have been journalled already"
This reverts commit b7ca305721.
2011-01-12 08:45:53 -05:00
Alberto Lerner
b7ca305721 fix: if durability is on, mods would have been journalled already 2011-01-11 10:13:30 -08:00
Eliot Horowitz
b828d21630 ran astyle SERVER-2304 2011-01-04 00:40:41 -05:00
Alberto Lerner
2e1be2738a correct messages 2010-12-29 11:59:31 -05:00
Alberto Lerner
6da107cc62 makes a bit easier to detect an aborted move in config.changelog 2010-12-28 18:49:53 -05:00
Alberto Lerner
5e5ed8bb03 message warns explicitly that operation aborted 2010-12-28 14:54:29 -05:00
Alberto Lerner
84a30c24e9 log if moveChunk/splitChunk did not pass admission checks 2010-12-27 14:30:54 -05:00
Alberto Lerner
6c3aac4b9a better logging messages 2010-12-19 17:55:36 -05:00
Alberto Lerner
fc38980536 SERVER-2040 deleted that message by mistake in last commit 2010-12-19 16:11:01 -05:00
Alberto Lerner
63c23a2359 SERVER-2024 if a network error is raised during migrate commit, pause, and recheck 2010-12-19 16:04:45 -05:00
Alberto Lerner
34613b0bde make finding a stuck/slow migrate a bit easier 2010-12-18 18:38:43 -05:00
Alberto Lerner
d3d11d6863 SERVER-1358 (Fix) chunk size accounting was using sum of all chunks on shards 2010-12-18 15:13:52 -05:00
Alberto Lerner
a3b3c83f96 SERVER-1358 don't migrate a chunk larger than max chunk size (but allow explicit override) 2010-12-17 17:49:47 -05:00
Alberto Lerner
2325d9c61a SERVER-2119 push migrated chunk data to secondary/journal at migrate commit 2010-12-15 17:19:51 -05:00
Alberto Lerner
4597deb9d4 SERVER-2024 disambiguate 2010-12-11 16:25:44 -05:00
Alberto Lerner
d2e9c85869 SERVER-2024 on migrate commit, either update both TO/FROM shards or none 2010-12-11 16:17:57 -05:00
Alberto Lerner
c6016d1ac7 fix: reset version to the right number if migrate commit fails 2010-12-08 16:16:29 -05:00
Alberto Lerner
b9c944466f hook new versioning machinery and allow mongod to maintain per-chunk state 2010-12-04 14:49:09 -05:00
Alberto Lerner
2382b52810 dblock is deprecated 2010-11-24 18:49:19 -05:00
Alberto Lerner
82dbe74c87 comment 2010-11-24 10:42:22 -05:00
Alberto Lerner
2478a98d79 SERVER-2119 Just to tie code site and JIRA item 2010-11-18 12:35:31 -05:00
Alberto Lerner
e58d6da404 SERVER-2024 reminder comment; implementation depends on new state on ChunkMater that's coming up 2010-11-12 23:41:41 -05:00
Alberto Lerner
e982a191b6 add 'passed' message 2010-11-09 17:38:08 -05:00
Alberto Lerner
8d306011ef SERVER-2068 make sure to use the same thread name 2010-11-08 09:41:35 -05:00
Eliot Horowitz
b741da2a0e when cleaning a chunk after migrating, don't store those deletes in the xfer list SERVER-2068 2010-11-08 00:17:09 -05:00
Alberto Lerner
f95f7a44c3 SERVER-1068 test deletions regardless of command line args 2010-11-07 18:12:05 -05:00
Alberto Lerner
0ede22a208 SERVER-2068 check on the TO side of a chunk migration before applying deletions 2010-11-07 16:31:53 -05:00
Alberto Lerner
5615486e98 SERVER-1974 rollback version if migrate aborts on the TO side 2010-10-26 09:52:50 -04:00
Eliot Horowitz
31984c056b during migrate, lock around each insertion to minimize impact on production systems 2010-10-24 01:52:44 -04:00
Eliot Horowitz
9923c7b636 yield when storing locs to migrate and yield while doing so 2010-10-23 17:40:58 -04:00
Alberto Lerner
abccecb42a MigrateStatus classes require thread-safe access 2010-10-23 11:23:58 -04:00
Alberto Lerner
d1ff952a41 SERVER-1979 check max key before accepting a move request 2010-10-20 16:32:49 -04:00
Alberto Lerner
cac8c78947 comment 2010-10-18 15:24:18 -04:00
Alberto Lerner
3e32ba710f release connection before returning 2010-10-18 12:13:11 -04:00
Alberto Lerner
ad76936753 extre semi-colon 2010-10-18 11:56:45 -04:00
Alberto Lerner
d344f95e6e which lock? 2010-10-18 11:02:08 -04:00
Eliot Horowitz
8b29ca04e5 use buffer len rather than computing ourself 2010-10-17 22:44:31 -04:00
Alberto Lerner
da51257c3b comments and readability 2010-10-12 13:57:24 -04:00
Eliot Horowitz
53a0d295e3 split bson max size into User and Internal
sometimes objects have to be bigger (oplog insertion for example)
more prep for SERVER-431
2010-10-11 11:26:41 -04:00
Eliot Horowitz
d3c3b8a903 using BSONObjMaxSize everywhere bson size comes into play
still 4mb, this is just prep for changing SERVER-431
2010-10-10 20:01:17 -04:00