Cherry-pick #20717 to 7.x: Improve registry file migration performance #20768
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
## What does this PR do? Ensure that fsync is called only once after the migration of old state entries is complete. ## Why is it important? The registry uses a checkpoint-predicate that when true writes all state to disk and calls fsync. The checkpoint operation is supposed to be disabled when migration the old registry file. During migration, the old state will be directly copied (after cleanup and schema changes are applied). The old state will only be deleted after the migration of all states is complete. Unfortunately, the checkpoint predicate did return true, instead of false, which did trigger a checkpoint operation per state to be migrated. The fix disables fsync, and now finalizes the migration by calling Checkpoint directly. The PR also provides a benchmark (each "op" is one migration attempt). Before this fix (go test did kill the run after 10min for 10k entries): ``` BenchmarkMigration0To1/1-32 286 4203353 ns/op BenchmarkMigration0To1/10-32 34 35730680 ns/op BenchmarkMigration0To1/100-32 2 720890839 ns/op BenchmarkMigration0To1/1000-32 1 31633569085 ns/op ... test timed out after 10min ``` Benchmark results with the fix (migration 100k entries took ~7.6s): ``` BenchmarkMigration0To1/1-32 274 4371400 ns/op BenchmarkMigration0To1/10-32 259 4639209 ns/op BenchmarkMigration0To1/100-32 100 13374147 ns/op BenchmarkMigration0To1/1000-32 13 104220944 ns/op BenchmarkMigration0To1/10000-32 2 916656798 ns/op BenchmarkMigration0To1/100000-32 1 7616648790 ns/op PASS ``` Closes elastic#20705 (cherry picked from commit 03748b3)
kvch
added
[zube]: In Review
backport
Team:Services
|
Pinging @elastic/integrations-services (Team:Services) |
botelastic
bot
added
needs_team
💔 Tests Failed
Expand to view the summary
Build stats
Test stats 🧪
Test errorsExpand to view the tests failures
Steps errorsExpand to view the steps failures
Log outputExpand to view the last 100 lines of log output
|
adriansr
approved these changes
Aug 25, 2020
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.
Cherry-pick of PR #20717 to 7.x branch. Original message:
What does this PR do?
Ensure that fsync is called only once after the migration of old state entries is complete.
Why is it important?
The registry uses a checkpoint-predicate that when true writes all state to
disk and calls fsync. The checkpoint operation is supposed to be disabled
when migration the old registry file.
During migration the old state will be directly copied (after cleanup
and schema changes are applied). The old state will only be deleted
after the migration of all states is complete.
Unfortunately the checkpoint predicate did return true, instead of
false, which did trigger a checkpoint operation per state to be
migrated. The fix disables fsync, and now finalizes the migration by
calling Checkpoint directly.
The PR also provides a benchmark (each "op" is one migration attempt). Before this fix (go test did kill the run after 10min for 10k entries):
Benchmark results with fix (migration 100k entries took ~7.6s):
Checklist
- [ ] I have made corresponding changes to the documentation- [ ] I have made corresponding change to the default configuration filesCHANGELOG.next.asciidocorCHANGELOG-developer.next.asciidoc.Related issues