changefeedccl: flush less #32060
Merged
changefeedccl: flush less #32060
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
For correctness, changeAggreagator needs to flush kafka before forwarding a span-level resolved timestamp to changeFrontier. Before this change, it was flushing once before every span-level resolved timestamp. On TPCC-1000, there were about 2000 of these per second, which meant we were spending more time flushing kafka than writing to it. (This is especially bad because it blocks Next calls.) This change batches span-level resolved timestamps for some amount of time, before flushing once and emitting them all. This adds ~20% delay to the changefeed-level timestamp from being emitted to the user slightly, but we now spend almost no time flushing. Along with cockroachdb#32023, optimistically: Closes cockroachdb#31001 Release note (bug fix): CHANGEFEEDs now spend dramatically less time flushing kafka writes
|
This change is |
mrtracy
approved these changes
Oct 31, 2018
|
TFTR bors r=mrtracy |
|
👎 Rejected by PR status |
|
bors r=mrtracy |
craig bot
pushed a commit
that referenced
this pull request
Oct 31, 2018
32060: changefeedccl: flush less r=mrtracy a=danhhz For correctness, changeAggreagator needs to flush kafka before forwarding a span-level resolved timestamp to changeFrontier. Before this change, it was flushing once before every span-level resolved timestamp. On TPCC-1000, there were about 2000 of these per second, which meant we were spending more time flushing kafka than writing to it. (This is especially bad because it blocks Next calls.) This change batches span-level resolved timestamps for some amount of time, before flushing once and emitting them all. This adds ~20% delay to the changefeed-level timestamp from being emitted to the user slightly, but we now spend almost no time flushing. Along with #32023, optimistically: Closes #31001 Release note (bug fix): CHANGEFEEDs now spend dramatically less time flushing kafka writes Co-authored-by: Daniel Harrison <[email protected]>
Build succeeded |
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.
For correctness, changeAggreagator needs to flush kafka before
forwarding a span-level resolved timestamp to changeFrontier. Before
this change, it was flushing once before every span-level resolved
timestamp. On TPCC-1000, there were about 2000 of these per second,
which meant we were spending more time flushing kafka than writing to
it. (This is especially bad because it blocks Next calls.)
This change batches span-level resolved timestamps for some amount of
time, before flushing once and emitting them all. This adds ~20% delay
to the changefeed-level timestamp from being emitted to the user
slightly, but we now spend almost no time flushing.
Along with #32023, optimistically:
Closes #31001
Release note (bug fix): CHANGEFEEDs now spend dramatically less time
flushing kafka writes