Move experimental frozen to frozen shard limit#71781
Move experimental frozen to frozen shard limit#71781henningandersen merged 10 commits intoelastic:masterfrom
Conversation
If master ends up on a newer version than other cluster members, we cannot apply the new index setting for shard limits. We skip doing so for now.
Frozen indices created on 7.12 would not belong to the frozen shard limit group, now we convert them when last node is upgraded. Relates elastic#71392
|
Pinging @elastic/es-distributed (Team:Distributed) |
| .put(DiskThresholdDecider.SETTING_IGNORE_DISK_WATERMARKS.getKey(), true); | ||
|
|
||
| // we cannot apply this setting during rolling upgrade. | ||
| // todo: update version after backport | ||
| if (minNodeVersion.onOrAfter(Version.V_8_0_0)) { | ||
| settings.put(ShardLimitValidator.INDEX_SETTING_SHARD_LIMIT_GROUP.getKey(), ShardLimitValidator.FROZEN_GROUP); | ||
| } |
|
Failure fixed by #71778 |
|
@elasticmachine run elasticsearch-ci/2 |
|
@elasticmachine update branch |
jasontedor
left a comment
There was a problem hiding this comment.
I left a few comments, but I don’t see a need for another round.
| return ClusterState.builder(currentState).metadata(builder).build(); | ||
| } | ||
|
|
||
| private static boolean wrongShardLimitGroup(org.elasticsearch.common.settings.Settings settings) { |
There was a problem hiding this comment.
This is a nit: how about notFrozenShardLimitGroup?
|
|
||
| @Override | ||
| public void onFailure(String source, Exception e) { | ||
| logger.warn("upgrading frozen indices to have frozen shard limit group failed", e); |
There was a problem hiding this comment.
maybe add , will retry on the next cluster state update?
There was a problem hiding this comment.
That is, we are letting the user know that something happened, at the warn level. Then they’re going to wonder what do they do? So let’s let them know, you don’t need to take any action, we’ll do it for you. Then, this is a purely informational error message from their perspective.
| } | ||
|
|
||
| public void initialize() { | ||
| clusterService.addListener(this::clusterChanged); |
There was a problem hiding this comment.
Maybe we can remove this listener after the upgrade succeeds?
| .put(indexMetadata.getSettings()) | ||
| .put(ShardLimitValidator.INDEX_SETTING_SHARD_LIMIT_GROUP.getKey(), ShardLimitValidator.FROZEN_GROUP) | ||
| ) | ||
| .settingsVersion(indexMetadata.getSettingsVersion() + 1) |
…sen/elasticsearch into fix_upgrade_shard_limit_pr2
|
@elasticmachine update branch |
|
@elasticmachine update branch |
Frozen indices created on 7.12 would not belong to the frozen shard limit group, now we convert them when last node is upgraded. Relates #71392
The versions to upgrade erroneously included until 8.0, now set to 7.13. Relates #71781
Frozen indices created on 7.12 would not belong to the frozen shard
limit group, now we convert them when last node is upgraded.
Relates #71392
This PR currently includes #71777, but that will be merged to master before this goes in.(done)Additionally, this PR has a few outstanding version related topics (todos), pending merge of #71760.