storage: Refactor block compaction to allow shard-splitting #2366
Conversation
This file contains 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
cyriltovena
commented
Sep 6, 2023
| sr := symbolsRewriter{ | ||
| profiles: it, | ||
| rewriters: make(map[BlockReader]*symdb.Rewriter, len(blocks)), | ||
| func newSymbolsRewriter(path string) *symbolsRewriter { |
There was a problem hiding this comment.
We now creates rewriters lazily as soon as I see the first row for a given block. Rewriters are not shared across multiple output blocks.
WDYT ?
There was a problem hiding this comment.
I like it :) Each block has to have its own symdb
|
Just need to finish writing some more robust tests then we should be good. I spotted a bug where start/end time is from original block and not actual profiles so I'll have to fix this. |
cyriltovena
commented
Sep 6, 2023
cyriltovena
commented
Sep 6, 2023
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Outdated suggestions cannot be applied.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 refactors the current compaction function
Compactto create a newCompactWithSplitting(...,shardCount uint64)function.The new function compacts and deduplicates data away from input blocks while splitting it into
shardCountshards new blocks. Some output blocks might be emtpy and not returned.Right now the shards key is the labels finguerprint (hash of all labels pair). I suspect we might want to add different sharding strategy in the future after some experiment.
To split into multiple blocks we don't hook into the reader anymore, we read row by row and push to the correct output block, turns out to be much cleaner. Specially since we don't have to write row groups ourselves anymore I discovered that if you configure correctly the parquet writer it will automatically handle flushing row groups.
I'm still working on adding more tests now.
PS: I suggest to review this in split view.