Commit Graph

1071 Commits

Author SHA1 Message Date
kimchy 5bd37f6f47 initial chunk of work on refactoring the gateway, syntatic gateway files names, commit points that include translog information 2010-08-23 14:48:49 +03:00
kimchy 1517fa3d28 initial chunk of work on refactoring the gateway, syntatic gateway files names, commit points that include translog information 2010-08-23 01:45:03 +03:00
kimchy 36ad3b246a addlow to add node allocation ot the module 2010-08-22 12:35:59 +03:00
kimchy 664f28492d improvements to the module system, allow for a module to spawn other modules, and pre process modules before injector creation 2010-08-22 10:42:25 +03:00
kimchy aa28b93610 refactor how throttling is done, instead of doing it after a shard is allocated to a node, and then wait till its allowed to recover, do it on the allocation level, and don't allocate a shard to a node that has N number of recoveries going on it 2010-08-22 02:47:34 +03:00
kimchy 7592862646 add a callback for allocation as well 2010-08-22 01:00:37 +03:00
kimchy a3852766ff only check for prefered allocation on data nodes 2010-08-22 00:32:43 +03:00
kimchy 3f701365b0 move the allocate replica after primary active logic to a node allocation 2010-08-22 00:14:56 +03:00
kimchy 5ded04c5cb refactor shard allocation to be more pluggable for node allocation "deciders" 2010-08-21 21:59:31 +03:00
kimchy 2865ceef85 more on sorted nodes when allocating shards above average number of shards per node 2010-08-21 19:55:38 +03:00
kimchy 76e1a6b1bf check also for negative values 2010-08-21 19:51:14 +03:00
kimchy 5d2c005622 pass to canAllocate (though nothing is done there now) the actual routing nodes 2010-08-21 17:52:24 +03:00
kimchy 1ddd969fa7 pass to canAllocate (though nothing is done there now) the actual routing nodes 2010-08-21 17:52:02 +03:00
kimchy 11dbbebb7a renamve var to better reflect its state (started, not active) 2010-08-21 17:50:08 +03:00
kimchy 3117341f44 Shards Allocation: Only rebalance a shard if all its instances are already active, closes #331. 2010-08-21 17:49:19 +03:00
kimchy 38e6649a7e Cluster Health API: Add `initializing_shards` and `unassigned_shards` to the response, closes #330. 2010-08-21 16:03:29 +03:00
kimchy 113ea1bb1b don't throttle gateway recovery based on peer recoveries (gateway recoveries are important to do as fast as possible). still throttle peer recoveries based on both. 2010-08-21 15:37:12 +03:00
kimchy e12bdd9faf rename RETRY stage in recovery to THROTTLE 2010-08-21 15:13:34 +03:00
kimchy 0c72bb2125 better writing of local md5 caches 2010-08-21 02:16:36 +03:00
kimchy d97695ad10 fix windows script to not double quote 2010-08-20 21:13:32 +03:00
kimchy 1ef1e9b25b don't spawn on received cluster state 2010-08-20 19:36:39 +03:00
kimchy 800e450e3a handle local md5 cases, don't include them in the list of files the directory has 2010-08-20 19:30:26 +03:00
kimchy b78597934f expose the maven params in the build files as sys envs 2010-08-20 17:50:04 +03:00
kimchy 9acae60dbf also have a minimum value for concurrent recoveries, at 3 2010-08-20 17:28:31 +03:00
kimchy f62f7b8ffe check the progress in case the index size has not been set yet... 2010-08-20 17:17:32 +03:00
kimchy 9f0d5f8ef1 better default concurrent recoveries, based on number of processors, tap it at 10 2010-08-20 17:06:56 +03:00
kimchy 8c25be6dee better exception message with primary not started exception on index/delete/delete_by_query apis 2010-08-20 16:36:25 +03:00
kimchy 07285d81d1 add stream done when there is no file (should not happen) 2010-08-20 15:59:46 +03:00
kimchy b609162be3 don't ignore recovery on throttling unless the shard is closed 2010-08-20 15:58:24 +03:00
kimchy 4e74001bde move marking recovering before creating a status 2010-08-20 15:24:49 +03:00
kimchy 4c00e89f5e move memory monitor benchamrk to single indexing stress test 2010-08-20 15:05:21 +03:00
kimchy ef64f5d970 move memory monitor benchamrk to single indexing stress test 2010-08-20 13:41:33 +03:00
kimchy f09b45c04c remove the changes file, annoying that its basically a copy of the issues 2010-08-20 13:16:07 +03:00
kimchy 58028aa9e5 tempate for changes 2010-08-20 13:04:25 +03:00
kimchy d79444d1b9 tempate for changes 2010-08-20 13:02:50 +03:00
kimchy c0552bdc70 Can't set "no stopwords" on analyzer, closes #329. 2010-08-20 03:22:43 +03:00
kimchy 6f1a9fbfb9 create md5 files on the local index storage as well, so they won't have to be recomputed when performing initial recovery 2010-08-20 02:28:14 +03:00
kimchy acae0650c8 bail out of recovery when close and still waiting on throttling 2010-08-20 01:24:02 +03:00
kimchy a05c367c5e don't snapshot gateway when cleaning shard 2010-08-20 00:59:37 +03:00
kimchy 4a2e076c6a do a reuse work prefetching not on the cluster state thread, so it won't block while fetching md5s from other nodes 2010-08-20 00:45:17 +03:00
kimchy a5c5f65da5 fix recovery status when all index got reused 2010-08-19 22:46:58 +03:00
kimchy 7437acfcea add progress to indices status API 2010-08-19 22:29:31 +03:00
kimchy 8d0e5b239c recator source rest building into common code 2010-08-19 20:24:12 +03:00
kimchy 7b1093fb16 anoter compressed string test 2010-08-19 18:45:15 +03:00
kimchy 19abe7a2a5 when using compressed source, extract the compressed source directly into the response without any buffering 2010-08-19 18:29:36 +03:00
kimchy 46ccee8f89 when deleting an index, remove any blocks associated with it 2010-08-19 17:21:14 +03:00
kimchy c40935ae14 store the mapping definition in compressed form internally to save memory 2010-08-19 17:06:36 +03:00
kimchy a2d10d490d allocate list when deserializing, NPE galore... 2010-08-19 16:40:51 +03:00
kimchy 04ae1581a0 fix NPE introduced on optimization of index status on docs status creation 2010-08-19 15:37:06 +03:00
kimchy 70c344b079 add compressed string 2010-08-19 15:33:57 +03:00