Fix SnapshotBasedRecoveryIT#testSnapshotBasedRecovery #77134
Conversation
Move the shard to a replica in an older version when the primary is located in the upgraded node during the first rolling restart round. Closes elastic#76595
fcofdez
added
>test
|
Pinging @elastic/es-distributed (Team:Distributed) |
henningandersen
left a comment
henningandersen
left a comment
There was a problem hiding this comment.
I left a couple of comments.
| String currentPrimaryNodeId = getPrimaryNodeIdOfShard(indexName, 0); | ||
| assertThat(getNodeVersion(currentPrimaryNodeId), is(equalTo(UPGRADE_FROM_VERSION))); | ||
|
|
||
| updateIndexSettings( |
There was a problem hiding this comment.
I think this risks balancing back the shard to the upgraded node. You can disable balancing and reenable it after reestablishing replica. Or we could just keep the exclusion until the next round. I think I prefer the latter.
| // Drop replicas | ||
| updateIndexSettings(indexName, Settings.builder().put(IndexMetadata.INDEX_NUMBER_OF_REPLICAS_SETTING.getKey(), 0)); | ||
| if (FIRST_MIXED_ROUND) { | ||
| String primaryNodeId = getPrimaryNodeIdOfShard(indexName, 0); |
There was a problem hiding this comment.
I wonder if we should not instead just find the node on the new version and put the exclusion in always? I think there is a risk of the shard moving between the check here and adding replicas below.
|
|
||
| // Drop replicas | ||
| updateIndexSettings(indexName, Settings.builder().put(IndexMetadata.INDEX_NUMBER_OF_REPLICAS_SETTING.getKey(), 0)); | ||
| ensureGreen(indexName); |
There was a problem hiding this comment.
I do not think this ensureGreen does anything?
| } | ||
| } | ||
|
|
||
| private void updateIndexSettings(String indexName, Settings settings) throws IOException { |
There was a problem hiding this comment.
This looks nearly identical to the static updateIndexSettings inherited from ESRestTestCase, do we need this?
|
@elasticmachine update branch |
|
@elasticmachine update branch |
fcofdez
added
the
auto-backport
💚 Backport successful
|
Move the shard to a replica in an older version when the primary is located in the upgraded node during the first rolling restart round. Closes elastic#76595
5 participants
Move the shard to a replica in an older version when the primary
is located in the upgraded node during the first rolling restart
round.
Closes #76595