Skip to content

Tags: magicmoux/elasticsearch

Tags

v5.4.3

Toggle v5.4.3's commit message
Elasticsearch v5.4.3

v5.4.2

Toggle v5.4.2's commit message
Elasticsearch v5.4.2

v6.0.0-alpha2

Toggle v6.0.0-alpha2's commit message
Elasticsearch v6.0.0-alpha2

v5.4.1

Toggle v5.4.1's commit message
Elasticsearch v5.4.1

v5.3.3

Toggle v5.3.3's commit message
Elasticsearch v5.3.3

v6.0.0-alpha1

Toggle v6.0.0-alpha1's commit message
Elasticsearch v6.0.0-alpha1

v5.4.0

Toggle v5.4.0's commit message
Elasticsearch v5.4.0

v5.3.2

Toggle v5.3.2's commit message
Only test hard linking to directory on macOS

This skips trying to create a hard link to a directory in the evil node
tests on non-macOS operating systems.

v5.3.1

Toggle v5.3.1's commit message
Elasticsearch v5.3.1

v2.4.5

Toggle v2.4.5's commit message
Fail shard marked as relocating but missing in-memory index structures (

elastic#24061)

When a shard is marked as relocating in the cluster state but the node that is to be expected to have the shard does not have the corresponding in-memory index structures loaded, then the node does not notify the master of this inconsistency. The cluster can end up in this situation for example when using discovery.zen.no_master_block: all, which makes the data nodes unload in-memory index structures when no master is available. As a result, all indexing / search activity to the supposedly relocating shard will fail, requiring manual intervention from an operator to fix the issue by restarting the data node with the relocating shard.