Hi, 

I just had this problem too after upgrading from 1.2.1 to 1.4.1
Basically, if your ES was upgraded recently to 1.4.x from <1.4.0, it's a 
very common 
situation https://github.com/elasticsearch/elasticsearch/issues/9922
In our case we just upgraded got corrupt shards, loaded snapshots, 
reindexed the delta information, got other broken shards on node restarts, 
and so on... until at some point, it just stopped happening.
We also reached the point where any snapshot failed (although it worked 
literally the day before).
Being insane 
(http://www.quotehd.com/imagequotes/TopAuthors/albert-einstein-physicist-insanity-doing-the-same-thing-over-and-over-again-and.jpg)
 
helped, because at some point the snap recovery worked.

I just had the issue again, with one shard stuck in "initializing" status 
when restarting the nodes after an upgrade.
Now that i've browsed hundreds of elastic search related documents, i know 
better :
Basically just identify the broken shard curl -XGET 
http://localhost:9200/_cluster/state?pretty=true > foo.json
The shard(s) will have "INITIALIZING" status and will have their node id 
associated, which will let you know where is the blocked shard.
If you're using one replica, and the other shard is efficiently started, 
you can just rm the directory containing the broken shard (the one you 
executer the check index thing) and it should rebuild itself based on the 
other, uncorrupted shard (from another node).

Now if you have a lot of corruption problems maybe take a look at this 
https://github.com/elasticsearch/elasticsearch/pull/7580 and at the output 
of your java -version.
Upgrading to java >= 1.7.55 is a requirement anyways directly from the 
elastic search website (can't remember where i've seen it exactly)

Also the check index tools does not repair anything it just destroys the 
index and creates a new one (from my experience).

It feels like your snapshot has nothing to do with so don't expect 
 anything from the AWS team.

And 1.4.0 and 1.4.1 have dreadful snapshot/restore and perm generation 
bugs, so i'd avoid these two.

What are your versions?

Hope this helped.

Le lundi 2 mars 2015 12:26:28 UTC+1, Sukanta Saha a écrit :
>
> Hi
>
> in all our elasticsearch cluster we use this elasticsearch-cloud-aws 
> plugin to create the snapshots on s3 on a regular basis.
>
> Some times we saw the shard got corrupted for an index in our 
> elasticsearch log.
> So we try to restore it from backup and while restoring it from backup 
> again we see the same exception in logs which is follows
>
> [2015-02-25 08:18:10,824][WARN ][indices.cluster ] 
> [test-es-cluster-1e-data-2] [lst_p113_v_4_20140615_0000][0] failed to start 
> shard
> org.elasticsearch.index.gateway.IndexShardGatewayRecoveryException: 
> [lst_p113_v_4_20140615_0000][0] failed recovery
> at 
> org.elasticsearch.index.gateway.IndexShardGatewayService$1.run(IndexShardGatewayService.java:185)
> at 
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> at 
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> at java.lang.Thread.run(Thread.java:745)
> Caused by: 
> org.elasticsearch.index.snapshots.IndexShardRestoreFailedException: 
> [lst_p113_v_4_20140615_0000][0] restore failed
> at 
> org.elasticsearch.index.snapshots.IndexShardSnapshotAndRestoreService.restore(IndexShardSnapshotAndRestoreService.java:130)
> at 
> org.elasticsearch.index.gateway.IndexShardGatewayService$1.run(IndexShardGatewayService.java:127)
> ... 3 more
> Caused by: 
> org.elasticsearch.index.snapshots.IndexShardRestoreFailedException: 
> [lst_p113_v_4_20140615_0000][0] failed to restore snapshot 
> [listening-prod6-20150224]
> at 
> org.elasticsearch.index.snapshots.blobstore.BlobStoreIndexShardRepository.restore(BlobStoreIndexShardRepository.java:165)
> at 
> org.elasticsearch.index.snapshots.IndexShardSnapshotAndRestoreService.restore(IndexShardSnapshotAndRestoreService.java:124)
> ... 4 more
> Caused by: 
> org.elasticsearch.index.snapshots.IndexShardRestoreFailedException: 
> [lst_p113_v_4_20140615_0000][0] Failed to recover index
> at 
> org.elasticsearch.index.snapshots.blobstore.BlobStoreIndexShardRepository$RestoreContext.restore(BlobStoreIndexShardRepository.java:787)
> at 
> org.elasticsearch.index.snapshots.blobstore.BlobStoreIndexShardRepository.restore(BlobStoreIndexShardRepository.java:162)
> ... 5 more
> Caused by: org.apache.lucene.index.CorruptIndexException: checksum failed 
> (hardware problem?) : expected=1lvsjli actual=3awj8p 
> resource=(org.apache.lucene.store.FSDirectory$FSIndexOutput@7266a49d)
> at 
> org.elasticsearch.index.store.LegacyVerification$Adler32VerifyingIndexOutput.verify(LegacyVerification.java:73)
> at org.elasticsearch.index.store.Store.verify(Store.java:365)
> at 
> org.elasticsearch.index.snapshots.blobstore.BlobStoreIndexShardRepository$RestoreContext.restoreFile(BlobStoreIndexShardRepository.java:843)
> at 
> org.elasticsearch.index.snapshots.blobstore.BlobStoreIndexShardRepository$RestoreContext.restore(BlobStoreIndexShardRepository.java:784)
> ... 6 more
> [2015-02-25 08:18:10,826][WARN ][cluster.action.shard ] 
> [test-es-cluster-1e-data-2] [lst_p113_v_4_20140615_0000][0] sending failed 
> shard for [lst_p113_v_4_20140615_0000][0], node[shNgLjr8RlW7Zrk3P4UdPg], 
> [P], restoring[aws-prod-elasticsearch-backup:listening-prod6-20150224], 
> s[INITIALIZING], indexUUID [ZQKQ-6naQqeLP1Gk8IFsig], reason [Failed to 
> start shard, message 
> [IndexShardGatewayRecoveryException[[lst_p113_v_4_20140615_0000][0] failed 
> recovery]; nested: 
> IndexShardRestoreFailedException[[lst_p113_v_4_20140615_0000][0] restore 
> failed]; nested: 
> IndexShardRestoreFailedException[[lst_p113_v_4_20140615_0000][0] failed to 
> restore snapshot [listening-prod6-20150224]]; nested: 
> IndexShardRestoreFailedException[[lst_p113_v_4_20140615_0000][0] Failed to 
> recover index]; nested: CorruptIndexException[checksum failed (hardware 
> problem?) : expected=1lvsjli actual=3awj8p 
> resource=(org.apache.lucene.store.FSDirectory$FSIndexOutput@7266a49d)]; ]]
>
> Even if we back to an older snapshot we found the same exception.
>
> So what we did was we download all the segments files from s3 merge it and 
> there we found some segments were corrupted by using
> org.apache.lucene.index.CheckIndex with -fix 
> We fixed it but we loose 5gb data.
>
> We shared this problem with elasticsearch-cloud-aws team , They didnot 
> give any reply till now,
>
> Can you guys please have a look into this issue and suggest something
>
> Thanks
>

-- 
You received this message because you are subscribed to the Google Groups 
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/elasticsearch/a9ea265d-aa70-4c2d-b117-0dd7e8b78ad6%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to