cockroach
cockroach copied to clipboard
Published
20 hours ago •
cockroachdb
cockroachdb
release-24.1: rangefeed: improve regression seen from memory accounting
Backport 1/1 commits from #123229 on behalf of @wenyihu6.
/cc @cockroachdb/release
This patch improves the regression we saw from #123076. Note that this pr only refactors the existing code but doesn't change any functionalities.
Related: https://github.com/cockroachdb/cockroach/issues/123076 Epic: none Release note: none
❯ benchdiff --new=a6e678fd7cb --old=13c9b9 --post-checkout='bash post-checkout.sh' --run=BenchmarkRangefeed --count=10 ./pkg/kv/kvserver/rangefeed
test binaries already exist for '13c9b9d'; skipping build
test binaries already exist for 'a6e678f'; skipping build
pkg=1/1 iter=10/10 cockroachdb/cockroach/pkg/kv/kvserver/rangefeed /
name old time/op new time/op delta
Rangefeed/budget=true/procType=scheduler/opType=closedts/numRegs=100-12 118µs ±14% 83µs ±62% -29.72% (p=0.002 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=commit/numRegs=100-12 131µs ±12% 100µs ±43% -23.78% (p=0.001 n=8+10)
Rangefeed/budget=false/procType=scheduler/opType=write/numRegs=100-12 119µs ±13% 102µs ±24% -13.96% (p=0.015 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=write/numRegs=1-12 561ns ±10% 524ns ± 3% -6.55% (p=0.001 n=10+8)
Rangefeed/budget=false/procType=scheduler/opType=closedts/numRegs=1-12 457ns ±10% 428ns ± 6% -6.47% (p=0.007 n=10+10)
Rangefeed/budget=false/procType=scheduler/opType=write/numRegs=1-12 513ns ± 3% 503ns ± 5% ~ (p=0.113 n=10+9)
Rangefeed/budget=false/procType=scheduler/opType=write/numRegs=10-12 6.15µs ±39% 5.41µs ±31% ~ (p=0.165 n=10+10)
Rangefeed/budget=false/procType=scheduler/opType=commit/numRegs=1-12 1.40µs ± 3% 1.40µs ± 7% ~ (p=1.000 n=9+10)
Rangefeed/budget=false/procType=scheduler/opType=commit/numRegs=10-12 6.07µs ±18% 6.15µs ±24% ~ (p=0.971 n=10+10)
Rangefeed/budget=false/procType=scheduler/opType=commit/numRegs=100-12 106µs ±29% 119µs ±10% ~ (p=0.077 n=9+9)
Rangefeed/budget=false/procType=scheduler/opType=closedts/numRegs=10-12 5.00µs ±41% 4.77µs ±39% ~ (p=0.670 n=10+10)
Rangefeed/budget=false/procType=scheduler/opType=closedts/numRegs=100-12 119µs ±17% 117µs ±25% ~ (p=0.604 n=10+9)
Rangefeed/budget=true/procType=scheduler/opType=write/numRegs=10-12 5.59µs ±22% 5.83µs ±30% ~ (p=0.481 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=write/numRegs=100-12 102µs ±60% 89µs ±62% ~ (p=0.631 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=commit/numRegs=1-12 1.63µs ± 4% 1.68µs ± 6% ~ (p=0.108 n=9+10)
Rangefeed/budget=true/procType=scheduler/opType=commit/numRegs=10-12 6.36µs ±18% 6.07µs ± 6% ~ (p=0.395 n=10+8)
Rangefeed/budget=true/procType=scheduler/opType=closedts/numRegs=1-12 477ns ± 7% 460ns ± 5% ~ (p=0.065 n=10+9)
Rangefeed/budget=true/procType=scheduler/opType=closedts/numRegs=10-12 5.08µs ±15% 5.24µs ±22% ~ (p=0.684 n=10+10)
RangefeedBudget/budget=false-12 533ns ± 7% 530ns ± 9% ~ (p=0.631 n=10+10)
RangefeedBudget/budget=true-12 546ns ± 8% 535ns ± 5% ~ (p=0.182 n=9+10)
name old alloc/op new alloc/op delta
Rangefeed/budget=true/procType=scheduler/opType=closedts/numRegs=1-12 340B ± 2% 262B ± 2% -22.79% (p=0.000 n=10+9)
Rangefeed/budget=false/procType=scheduler/opType=closedts/numRegs=1-12 359B ± 2% 285B ± 2% -20.61% (p=0.000 n=9+10)
Rangefeed/budget=false/procType=scheduler/opType=closedts/numRegs=10-12 414B ± 2% 332B ± 2% -19.81% (p=0.000 n=10+9)
Rangefeed/budget=false/procType=scheduler/opType=commit/numRegs=100-12 957B ± 3% 782B ± 4% -18.29% (p=0.000 n=9+9)
Rangefeed/budget=false/procType=scheduler/opType=closedts/numRegs=100-12 439B ± 5% 359B ±17% -18.28% (p=0.000 n=9+10)
Rangefeed/budget=false/procType=scheduler/opType=commit/numRegs=10-12 932B ± 4% 764B ± 6% -18.01% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=closedts/numRegs=10-12 452B ± 2% 374B ± 2% -17.27% (p=0.000 n=10+10)
Rangefeed/budget=false/procType=scheduler/opType=commit/numRegs=1-12 927B ± 4% 774B ± 4% -16.56% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=write/numRegs=1-12 473B ± 3% 395B ± 1% -16.34% (p=0.000 n=10+8)
Rangefeed/budget=true/procType=scheduler/opType=commit/numRegs=10-12 955B ± 1% 806B ± 5% -15.65% (p=0.000 n=8+10)
Rangefeed/budget=true/procType=scheduler/opType=commit/numRegs=1-12 944B ± 4% 797B ± 5% -15.61% (p=0.000 n=10+10)
RangefeedBudget/budget=true-12 469B ± 1% 397B ± 1% -15.39% (p=0.000 n=9+10)
Rangefeed/budget=false/procType=scheduler/opType=write/numRegs=1-12 488B ± 1% 413B ± 1% -15.37% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=commit/numRegs=100-12 1.01kB ± 5% 0.85kB ± 8% -15.31% (p=0.000 n=10+10)
RangefeedBudget/budget=false-12 490B ± 1% 416B ± 2% -15.18% (p=0.000 n=10+10)
Rangefeed/budget=false/procType=scheduler/opType=write/numRegs=10-12 539B ± 1% 459B ± 3% -14.86% (p=0.000 n=8+9)
Rangefeed/budget=true/procType=scheduler/opType=write/numRegs=10-12 584B ± 2% 502B ± 1% -14.11% (p=0.000 n=9+9)
Rangefeed/budget=true/procType=scheduler/opType=closedts/numRegs=100-12 480B ± 8% 418B ±13% -12.94% (p=0.001 n=9+10)
Rangefeed/budget=false/procType=scheduler/opType=write/numRegs=100-12 555B ± 1% 486B ± 4% -12.47% (p=0.000 n=8+10)
Rangefeed/budget=true/procType=scheduler/opType=write/numRegs=100-12 601B ± 5% 534B ± 4% -11.15% (p=0.000 n=8+9)
name old allocs/op new allocs/op delta
Rangefeed/budget=false/procType=scheduler/opType=closedts/numRegs=10-12 6.60 ± 9% 5.00 ± 0% -24.24% (p=0.000 n=10+8)
Rangefeed/budget=false/procType=scheduler/opType=commit/numRegs=10-12 12.0 ± 0% 10.0 ± 0% -16.67% (p=0.000 n=10+10)
Rangefeed/budget=false/procType=scheduler/opType=commit/numRegs=100-12 12.0 ± 0% 10.0 ± 0% -16.67% (p=0.000 n=10+10)
Rangefeed/budget=false/procType=scheduler/opType=closedts/numRegs=1-12 6.00 ± 0% 5.00 ± 0% -16.67% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=commit/numRegs=1-12 12.0 ± 0% 10.0 ± 0% -16.67% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=closedts/numRegs=1-12 6.00 ± 0% 5.00 ± 0% -16.67% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=commit/numRegs=100-12 13.0 ± 0% 11.0 ± 0% -15.38% (p=0.000 n=10+10)
Rangefeed/budget=false/procType=scheduler/opType=commit/numRegs=1-12 11.0 ± 0% 9.4 ± 6% -14.55% (p=0.000 n=9+10)
Rangefeed/budget=false/procType=scheduler/opType=closedts/numRegs=100-12 7.00 ± 0% 6.00 ± 0% -14.29% (p=0.000 n=9+10)
Rangefeed/budget=true/procType=scheduler/opType=closedts/numRegs=10-12 7.00 ± 0% 6.00 ± 0% -14.29% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=commit/numRegs=10-12 12.0 ± 0% 10.3 ± 7% -14.17% (p=0.000 n=9+10)
Rangefeed/budget=false/procType=scheduler/opType=write/numRegs=1-12 8.00 ± 0% 7.00 ± 0% -12.50% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=write/numRegs=1-12 8.00 ± 0% 7.00 ± 0% -12.50% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=closedts/numRegs=100-12 8.00 ± 0% 7.00 ± 0% -12.50% (p=0.000 n=7+10)
RangefeedBudget/budget=false-12 8.00 ± 0% 7.00 ± 0% -12.50% (p=0.000 n=10+10)
RangefeedBudget/budget=true-12 8.00 ± 0% 7.00 ± 0% -12.50% (p=0.000 n=10+10)
Rangefeed/budget=false/procType=scheduler/opType=write/numRegs=100-12 9.00 ± 0% 8.00 ± 0% -11.11% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=write/numRegs=10-12 9.00 ± 0% 8.00 ± 0% -11.11% (p=0.000 n=10+10)
Rangefeed/budget=true/procType=scheduler/opType=write/numRegs=100-12 10.0 ± 0% 9.0 ± 0% -10.00% (p=0.000 n=9+10)
Rangefeed/budget=false/procType=scheduler/opType=write/numRegs=10-12 8.00 ± 0% 7.30 ±10% -8.75% (p=0.001 n=8+10)
Release justification: low risk improvement on 24.1 regression
![blathers-crl[bot] avatar](https://avatars.githubusercontent.com/in/59700?v=4 "Published by a pub.dev verified publisher"){kind=small}
Thanks for opening a backport.
Please check the backport criteria before merging:
- [x] Backports should only be created for serious issues or test-only changes.
- [x] Backports should not break backwards-compatibility.
- [x] Backports should change as little code as possible.
- [x] Backports should not change on-disk formats or node communication protocols.
- [x] Backports should not add new functionality (except as defined here).
- [x] Backports must not add, edit, or otherwise modify cluster versions; or add version gates.
- [ ] All backports must be reviewed by the owning areas TL and one additional TL. For more information as to how that review should be conducted, please consult the backport policy.
If your backport adds new functionality, please ensure that the following additional criteria are satisfied:
- [ ] There is a high priority need for the functionality that cannot wait until the next release and is difficult to address in another way.
- [ ] The new functionality is additive-only and only runs for clusters which have specifically “opted in” to it (e.g. by a cluster setting).
- [ ] New code is protected by a conditional check that is trivial to verify and ensures that it only runs for opt-in clusters. State changes must be further protected such that nodes running old binaries will not be negatively impacted by the new state (with a mixed version test added).
- [ ] The PM and TL on the team that owns the changed code have signed off that the change obeys the above rules.
- [ ] Your backport must be accompanied by a post to the appropriate Slack channel (#db-backports-point-releases or #db-backports-XX-X-release) for awareness and discussion.
Also, please add a brief release justification to the body of your PR to justify this backport.
I'm not 100% sure if this is backportable. Happy to close it if we don't feel comfortable backporting it.
