fix(blooms): Fully deduplicate chunks from FilterChunkRef responses #12807

chaudum · 2024-04-26T19:53:14Z

What this PR does / why we need it:

This PR aims for full de-duplication of chunks and series from filter requests from the index gateway to the bloom gateway.

Whenever we merge/de-duplicate slices, the inputs need to be sorted. It appears that the Removals (chunks) from the v1.Output are not guaranteed to be sorted.

When comparing ShortRefs, both From, Through, and Checksum need to be used.

Checklist

Reviewed the CONTRIBUTING.md guide (required)
Documentation added
Tests updated
Title matches the required conventional commits format, see here
Changes that require user attention or interaction to upgrade are documented in docs/sources/setup/upgrade/_index.md
For Helm chart changes bump the Helm chart version in production/helm/loki/Chart.yaml and update production/helm/loki/CHANGELOG.md and production/helm/loki/README.md. Example PR
If the change is deprecating or removing a configuration option, update the deprecated-config.yaml and deleted-config.yaml files respectively in the tools/deprecated-config-checker directory. Example PR

Signed-off-by: Christian Haudum <[email protected]>

for the case when chunks have same from/through time, but different checksums Signed-off-by: Christian Haudum <[email protected]>

Signed-off-by: Christian Haudum <[email protected]>

grafanabot · 2024-04-30T06:14:18Z

This PR must be merged before a backport PR will be created.

grafanabot · 2024-04-30T06:14:27Z

This PR must be merged before a backport PR will be created.

salvacorts · 2024-04-30T08:04:32Z

pkg/bloomgateway/bloomgateway.go

@@ -383,6 +383,8 @@ func (g *Gateway) consumeTask(ctx context.Context, task Task, tasksCh chan<- Tas
 		case <-ctx.Done():
 			// do nothing
 		default:
+			// chunks may not be sorted
+			sort.Slice(res.Removals, func(i, j int) bool { return res.Removals[i].Less(res.Removals[j]) })


nit: We tend to use sort very often and I'm worried we end up sorting already sorted slices (quicksort can degrade to O(n^2) for sorted inputs). Should we check if the list is sorted before sorting it?

good point.

afaik, pdqsort which is used by sort.Slice() prevents the O(n^2) worst case scenario from quicksort and can do O(n log n) instead (which is still bad).

salvacorts · 2024-04-30T08:10:52Z

pkg/bloomgateway/client.go

+		if a.Less(b) {
+			result = append(result, a)
+			i++
+		} else {


nite: more idiomatic to add continue inside the if than having an else.

Signed-off-by: Christian Haudum <[email protected]>

salvacorts

Approving but leaving a nit here

salvacorts · 2024-04-30T10:49:45Z

pkg/bloomgateway/bloomgateway.go

-			// chunks may not be sorted
-			sort.Slice(res.Removals, func(i, j int) bool { return res.Removals[i].Less(res.Removals[j]) })
+			// chunks may not always be sorted
+			if !slices.IsSortedFunc(res.Removals, func(a, b v1.ChunkRef) int { return a.Cmp(b) }) {


nit: we end up using the same cmp funtion all over the places. Would be nice to have another function next to func (r *ChunkRef) Cmp(other ChunkRef) int that does this:

func CmpChunkRefs(a, b ChunkRef) int { return a.cmp(b) }

So we can use IsSortedFunc and SortFunc as:

if !slices.IsSortedFunc(a.Refs, CmpChunkRefs) { slices.SortFunc(a.Refs, CmpChunkRefs) } if !slices.IsSortedFunc(b.Refs, CmpChunkRefs) { slices.SortFunc(b.Refs, CmpChunkRefs) }

…12807) This PR aims for full de-duplication of chunks and series from filter requests from the index gateway to the bloom gateway. Whenever we merge/de-duplicate slices, the inputs need to be sorted. It appears that the Removals (chunks) from the v1.Output are not guaranteed to be sorted. When comparing ShortRefs, both From, Through, and Checksum need to be used. Signed-off-by: Christian Haudum <[email protected]> (cherry picked from commit a0f358f)

…ef responses (#12835) Backport a0f358f from #12807 Co-authored-by: Christian Haudum <[email protected]>

…12807) This PR aims for full de-duplication of chunks and series from filter requests from the index gateway to the bloom gateway. Whenever we merge/de-duplicate slices, the inputs need to be sorted. It appears that the Removals (chunks) from the v1.Output are not guaranteed to be sorted. When comparing ShortRefs, both From, Through, and Checksum need to be used. Signed-off-by: Christian Haudum <[email protected]>

pull-request-size bot added the size/M label Apr 26, 2024

chaudum force-pushed the chaudum/full-dedupe branch from 2a084f3 to 3d4c66c Compare April 29, 2024 07:41

pull-request-size bot added size/L and removed size/M labels Apr 29, 2024

chaudum force-pushed the chaudum/full-dedupe branch from 3281995 to 95b011b Compare April 29, 2024 19:22

chaudum marked this pull request as ready for review April 29, 2024 19:52

chaudum requested a review from a team as a code owner April 29, 2024 19:52

chaudum added 6 commits April 30, 2024 08:12

Dedupe series that overlap multiple days

2d80694

Signed-off-by: Christian Haudum <[email protected]>

Fix chunk dedupe iterator

2c9be4d

for the case when chunks have same from/through time, but different checksums Signed-off-by: Christian Haudum <[email protected]>

Cleanup

a0f4c2d

Signed-off-by: Christian Haudum <[email protected]>

Remove excessive logging

2cc43f2

Signed-off-by: Christian Haudum <[email protected]>

Don't overallocate

52ec607

Signed-off-by: Christian Haudum <[email protected]>

Add comment

628f1c2

Signed-off-by: Christian Haudum <[email protected]>

chaudum force-pushed the chaudum/full-dedupe branch from 75517c9 to 628f1c2 Compare April 30, 2024 06:13

chaudum added type/bug Somehing is not working as expected backport k200 labels Apr 30, 2024

salvacorts reviewed Apr 30, 2024

View reviewed changes

chaudum added 3 commits April 30, 2024 11:19

More idiomatic go

2dda1e8

Signed-off-by: Christian Haudum <[email protected]>

Check if sorted before sorting

1a77702

Signed-off-by: Christian Haudum <[email protected]>

fixup! Check if sorted before sorting

4f5837e

Signed-off-by: Christian Haudum <[email protected]>

salvacorts approved these changes Apr 30, 2024

View reviewed changes

chaudum merged commit a0f358f into main Apr 30, 2024
59 checks passed

chaudum deleted the chaudum/full-dedupe branch April 30, 2024 11:33

grafanabot mentioned this pull request Apr 30, 2024

chore: [k200] fix(blooms): Fully deduplicate chunks from FilterChunkRef responses #12835

Merged

7 tasks

chaudum added a commit that referenced this pull request Apr 30, 2024

chore: [k200] fix(blooms): Fully deduplicate chunks from FilterChunkR…

9e9852c

…ef responses (#12835) Backport a0f358f from #12807 Co-authored-by: Christian Haudum <[email protected]>

chaudum added a commit that referenced this pull request May 3, 2024

chore: [k200] fix(blooms): Fully deduplicate chunks from FilterChunkR…

8785e74

…ef responses (#12835) Backport a0f358f from #12807 Co-authored-by: Christian Haudum <[email protected]>

This was referenced May 6, 2024

chore(k201): release 3.1.0 #12894

Closed

chore(k202): release 3.1.0 #12945

Closed

loki-gh-app bot mentioned this pull request May 20, 2024

chore(k203): release 3.1.0 #12988

Closed

This was referenced May 27, 2024

chore(k204): release 3.1.0 #13037

Closed

chore(k205): release 3.1.0 #13102

Closed

This was referenced Jun 10, 2024

chore(k206): release 3.1.0 #13184

Closed

chore(k207): release 3.1.0 #13225

Merged

loki-gh-app bot mentioned this pull request Jun 24, 2024

chore(k208): release 3.1.0 #13291

Closed

loki-gh-app bot mentioned this pull request Jul 1, 2024

chore(k209): release 3.1.0 #13356

Closed

grafanabot mentioned this pull request Jul 2, 2024

chore: [main] chore(k207): release 3.1.0 #13391

Open

This was referenced Jul 3, 2024

chore(release-3.1.x): release 3.0.1 #13402

Closed

chore(k210): release 3.1.0 #13435

Closed

chore(k210): release 3.1.0 #13462

Closed

RodrigoCMoraes mentioned this pull request Jul 12, 2024

incognia inloco/loki#22

Closed

loki-gh-app bot mentioned this pull request Jul 15, 2024

chore(k211): release 3.1.0 #13521

Closed

loki-gh-app bot mentioned this pull request Jul 22, 2024

chore(k212): release 3.1.0 #13595

Closed

loki-gh-app bot mentioned this pull request Oct 18, 2024

chore(release-3.1.x): release 3.1.3 #14531

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(blooms): Fully deduplicate chunks from FilterChunkRef responses #12807

fix(blooms): Fully deduplicate chunks from FilterChunkRef responses #12807

chaudum commented Apr 26, 2024 •

edited

Loading

grafanabot commented Apr 30, 2024

grafanabot commented Apr 30, 2024

salvacorts Apr 30, 2024

chaudum Apr 30, 2024

salvacorts Apr 30, 2024

salvacorts left a comment

salvacorts Apr 30, 2024

fix(blooms): Fully deduplicate chunks from FilterChunkRef responses #12807

fix(blooms): Fully deduplicate chunks from FilterChunkRef responses #12807

Conversation

chaudum commented Apr 26, 2024 • edited Loading

grafanabot commented Apr 30, 2024

grafanabot commented Apr 30, 2024

salvacorts Apr 30, 2024

Choose a reason for hiding this comment

chaudum Apr 30, 2024

Choose a reason for hiding this comment

salvacorts Apr 30, 2024

Choose a reason for hiding this comment

salvacorts left a comment

Choose a reason for hiding this comment

salvacorts Apr 30, 2024

Choose a reason for hiding this comment

chaudum commented Apr 26, 2024 •

edited

Loading