[WIP] Various Speedups to Cluster State Updates (ILM + Allocation) #79692
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…pt-auto-expand-replicas
…r-aadapt-auto-expand-replicas
original-brownbear
removed
WIP
:Distributed Coordination/Allocation
|
closing here, I merged what I wanted from this to mainline now |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This contains a number of fixes/speedups to the allocation code as well as to ILM. This will be broken down into multiple pieces, some of which need quite a bit of polishing still.
That said, this branch speeds up certain cluster state updates massively. E.g. rolling over a data stream when there's already 30k indices present in a cluster goes from ~5s in our benchmark environment to ~1s.
The fixes to ILM in this PR effectively speed up every cluster state update, the larger the cluster the more the speedup and make ILM close to a noop as long as there's nothing for it to execute.
Memory consumption is also reduced to less than 1G of heap for 30k indices, from ~4G of heap in current 7.16 for the same cluster state.
I will start breaking things out of this PR one by one, opening PRs to master. This draft is mainly for keeping track of the work in in and some discussion.