-
Notifications
You must be signed in to change notification settings - Fork 3.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
feat(wal): Benchmark and improve WAL writes using Reset. #13272
Conversation
pkg/storage/wal/segment.go
Outdated
streams: swiss.NewMap[streamID, *streamSegment](64), | ||
buf1: encoding.EncWith(make([]byte, 0, 4)), | ||
func NewWalSegmentWriter() (*SegmentWriter, error) { | ||
idxWriter, err := index.NewWriter(context.TODO()) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we have to pass a ctx to the index? I think we can remove it because we only ever pass context.TODO()
. We also don't pass a context to the parent SegmentWriter so the utility is very limited.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes I'll look into that.
pkg/storage/wal/segment.go
Outdated
s = streamSegmentPool.Get().(*streamSegment) | ||
s.lbls = lbls | ||
s.tenantID = tenantID | ||
s.entries = s.entries[:0] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should a streamSegment have a reset()
method?
It looks like only one line to reset them but it would maintain consistency with our other Readers/Writers.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM! Just one comment on the tests but the benchmark should already cover it so happy to approve regardless.
t.Logf("Series sizes: [%s]\n", sizesString) | ||
} | ||
|
||
func TestReset(t *testing.T) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Could you add/modify this test case so the re-use case adds more data than the initial case?
If we had re-use errors, we'd likely panic when re-using a small buffer buit we'd be less likely to panic when re-using a large buffer.
for i := 0; i < b.N; i++ { | ||
writer := pool.Get().(*SegmentWriter) | ||
|
||
dst.Reset() |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This benchmark isn't running in parallel (since dst is reused). You could remove the pool for writers from the benchmark, or add a pool for dst and run the whole thing in parallel :)
What this PR does / why we need it:
This adds a benchmark and some improvement to the write path of wal segment.
Which issue(s) this PR fixes:
Fixes https://github.com/grafana/loki-private/issues/1005 https://github.com/grafana/loki-private/issues/1004