Commit Graph

24251 Commits

Author SHA1 Message Date
kimchy b657ffc5e7 Zen Discovery: Control which nodes are allowed to become masters, closes #248. 2010-07-09 04:22:03 +03:00
kimchy d1f5577c40 add channel based read from fs snapshot, reusing the same file descriptor with direct position based reads 2010-07-09 03:51:36 +03:00
kimchy 6cdd6a015c reuse file buffer 2010-07-09 03:24:31 +03:00
kimchy d531d82cfb Zen Discovery: A node might get into an infinite state of trying to find a master (when client / non_master) nodes exists, closes #247. 2010-07-09 00:43:06 +03:00
kimchy 87eb6bed8f nicer logging 2010-07-08 22:53:41 +03:00
kimchy 720e6a6d5b nicer logging 2010-07-08 22:52:03 +03:00
kimchy be6aaa157f fixed a bug in fs translog where it wasn't seeking correctly, and not counting the number of operations 2010-07-08 22:18:26 +03:00
kimchy ef148077fc better logging 2010-07-08 21:37:10 +03:00
kimchy 84af9f5681 improve handling of translog snapshotting 2010-07-08 20:51:37 +03:00
kimchy 4429a61528 better cleanup on shard startup failure 2010-07-08 19:20:03 +03:00
kimchy b69fc265c5 index writer ram buffer size can only be 2G, make sure not to set an illegal value 2010-07-08 19:03:55 +03:00
kimchy 400ec7b8d1 index writer ram buffer size can only be 2G, make sure not to set an illegal value 2010-07-08 18:58:04 +03:00
kimchy 63deef42cf add a help appendable blob store based on immutable one 2010-07-08 18:53:40 +03:00
kimchy 8d669ff54e move sorting to be done using the new fielddata cache, and not using lucene built in FieldCache 2010-07-08 14:42:35 +03:00
kimchy 97b5b41522 enable eager cache eviction based on readers, requires a nasty hack to Lucene, basically overriding two classes to hook into when a reader actually gets closed (and only enabled when elasticsearch is before lucene in the classpath, which we take for in the startup scripts) 2010-07-07 17:15:26 +03:00
kimchy a36b06c682 ignore failed snapshot on a non started index 2010-07-07 16:59:33 +03:00
kimchy a8c317e355 fix terms facets test to take into account distributed terms request 2010-07-07 16:33:39 +03:00
kimchy 5ff46911b6 ignore recovery when no node to recover from (it has disconnected, for example) 2010-07-07 16:30:28 +03:00
kimchy 3fb68c52da ignore snapshot on a shard thas has just been created 2010-07-07 16:30:17 +03:00
kimchy 11a8d637c0 ignore snapshot on a shard thas has just been created 2010-07-07 16:26:19 +03:00
kimchy e40aa540fe ignore recovery when no node to recover from (it has disconnected, for example) 2010-07-07 16:25:55 +03:00
kimchy 9578ad3ef1 Terms Facets: Allow to specify a set of terms to exclude in the request, closes #246. 2010-07-07 14:41:34 +03:00
kimchy 2b5458daf6 implement a simple file system based translog, not the defeault yet, should provide better memory management 2010-07-07 09:30:07 +03:00
kimchy b2bdb149f9 log shard close failure when closing an index 2010-07-06 22:50:40 +03:00
kimchy 6aa9be238c use the atomic integer counter to represent the size 2010-07-06 18:45:19 +03:00
kimchy 267016758d improve handling of memory caching with file system, only force compound file when really needed (when an extension that exists within the compound file is part of the memory cached extensions) 2010-07-06 18:21:36 +03:00
kimchy ac01bb86e8 clean lock only before opening an index writer 2010-07-06 18:13:05 +03:00
kimchy 2908418d37 checn the default byte buffer cache to 200mb, now that its global per node 2010-07-06 17:58:30 +03:00
kimchy 070cb5b295 allocate replica shards (backups) only after the primaries have been allocated, this results in better theoretical shard allocation, and the ability to reuse local storage index files (comared against the backup) 2010-07-06 17:58:05 +03:00
kimchy 1d39bb4d51 cache md5 computation for unallocated shard and reused it 2010-07-06 09:00:24 +03:00
kimchy d4f86899e3 more work on reuse work dir, improved transaction log to move to stream based 2010-07-06 00:12:40 +03:00
kimchy b078c9206a more work on reuse work dir, added md5 logic to identify what need to be recovered or not, note, requires reindexing 2010-07-05 14:13:55 +03:00
kimchy 79b69764e0 doc 2010-07-01 15:03:11 +03:00
kimchy e8b261c165 better shutdown handling 2010-07-01 10:28:57 +03:00
kimchy ad5525f575 fix unwrapping of exceptions 2010-07-01 10:27:53 +03:00
kimchy f6c9c49645 better logging when failing to read full content of a transport message 2010-07-01 03:00:57 +03:00
kimchy c75a56ca17 better shutdown handling 2010-07-01 02:41:00 +03:00
kimchy c2d02e4e3a wrong serialization of shardId 2010-07-01 02:18:35 +03:00
kimchy d848ab3df3 rename executors to esexecutors 2010-07-01 00:06:34 +03:00
kimchy b0d736b489 make the shard routing an actual component 2010-06-30 23:39:27 +03:00
kimchy 811559f393 add another test if the pool size is greater than the core pool size before a worker can exit 2010-06-30 23:08:05 +03:00
kimchy 6709518c4e add the ability to list shard index blobs without actually having one created 2010-06-30 22:54:59 +03:00
kimchy dea8238ced Move service wrapper support to its own repo, closes #243. 2010-06-30 21:51:34 +03:00
kimchy 66096e97f2 Remove the terms API (replaced by facets in search), closes #242. 2010-06-30 19:49:41 +03:00
kimchy 53b66fb26a revert committed configuration 2010-06-30 16:11:41 +03:00
kimchy df7e3f55a4 improve cluster blocks data structures for simpler usage 2010-06-30 16:06:31 +03:00
kimchy 4049da97fb time value to support parsing days as well 2010-06-30 13:00:25 +03:00
kimchy 088e0b5a64 Automatic management of indexing buffer size, closes #241. 2010-06-30 01:46:57 +03:00
kimchy fdb2eff998 Automatic management of indexing buffer size, closes #241. 2010-06-29 22:54:54 +03:00
kimchy 03ac38fdba better benchmark on memory monitor 2010-06-29 22:12:43 +03:00