Elastic shard corrupted and unassigned

[fernandamaffioletti (fernanda)]

Hi! Our elasticsearch stoped allocating one shard.

This is the error:
[2019-09-18 14:41:03,707][WARN ][cluster.action.shard ] [DMESPROD3] [dmesprod][0] received shard failed for [dmesprod][0], node[OcRTFaolS9mX6Cni7gWtGw], [P], s[INITIALIZING], indexUUID [fSnPdqL_S2CMIgUSagQO-A], reason [Failed to start shard, message [IndexShardGatewayRecoveryException[[dmesprod][0] failed to fetch index version after copying it over]; nested: CorruptIndexException[[dmesprod][0] Corrupted index [corrupted_xaX-iT4cQtCjGdDQG-w1Zg] caused by: CorruptIndexException[codec footer mismatch: actual footer=272697361 vs expected footer=-1071082520 (resource: NIOFSIndexInput(path="/var/lib/elasticsearch/dmes/nodes/0/indices/dmesprod/0/index/_1yake_es090_0.doc"))]]; ]]

how can we recover this shard?

[DavidTurner]

Hi @fernandamaffioletti and welcome!

This shard is corrupt and needs to be rebuilt from a good copy on a different node. Is your cluster health red? If not, Elasticsearch has already rebuilt it from a replica. If it is red, can you use the allocation explain API to find out why Elasticsearch cannot allocate any of the other copies of this shard either? You can share the output here if you need help interpreting it:

GET /_cluster/allocation/explain
{
  "index": "dmesprod",
  "shard": 0,
  "primary": true
}