feat: allow list items to be processed in parallel #738
Conversation
This file contains hidden or 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
shady-canva
force-pushed
the
shady-parallel-list-item-processing
branch
2 times, most recently
from
849c153 to
0f1cd01
Compare
Signed-off-by: Shady Rafehi <[email protected]>
shady-canva
force-pushed
the
shady-parallel-list-item-processing
branch
from
0f1cd01 to
4e4d254
Compare
|
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #738 +/- ##
==========================================
- Coverage 54.26% 47.67% -6.59%
==========================================
Files 64 64
Lines 6164 6611 +447
==========================================
- Hits 3345 3152 -193
- Misses 2549 3199 +650
+ Partials 270 260 -10 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
| listPageSize: defaultListPageSize, | ||
| listPageBufferSize: defaultListPageBufferSize, | ||
| listSemaphore: semaphore.NewWeighted(defaultListSemaphoreWeight), | ||
| listItemSemaphoreWeight: defaultListItemSemaphoreWeight, |
There was a problem hiding this comment.
this is great that it maintains existing behavior by default
Comment on lines
678
to
694
| resourceVersion, err := c.listResources(ctx, resClient, func(listPager *pager.ListPager) error { | ||
| return listPager.EachListItem(ctx, metav1.ListOptions{}, func(obj runtime.Object) error { | ||
| if un, ok := obj.(*unstructured.Unstructured); !ok { | ||
| return fmt.Errorf("object %s/%s has an unexpected type", un.GroupVersionKind().String(), un.GetName()) | ||
| } else { | ||
| items = append(items, c.newResource(un)) | ||
| if err := limiter.Run(ctx, func() { | ||
| newRes := c.newResource(un) | ||
| listLock.Lock() | ||
| items = append(items, newRes) | ||
| listLock.Unlock() | ||
| }); err != nil { | ||
| return fmt.Errorf("failed to process list item: %w", err) | ||
| } | ||
| } | ||
| return nil | ||
| }) | ||
| }) |
There was a problem hiding this comment.
Just to call this out as a trade-off, I think this will cause a change in ordering when using concurrency. But this is only when its enabled. Not a bad trade-off, just something to be aware of.
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.
This PR introduces a new setting which allows items returned by
ListPagerto be processed in parallel.Currently, items are processed sequentially per resource. This is a problem if
populateResourceInfoHandleris expensive. We internally have a custom resource which takes 1.5 milliseconds in thepopulateResourceInfoHandlerfunction due to a custom lua healthcheck script and a few ignored labels which are removed during diff normalisation. With over 50,000 of these custom resources, that's a 75 second overhead.With a parallelism of 8, we've seen a page of 500 items go from ~750ms to ~100ms. We've also seen the processing of a full list complete at roughly the same time as
ListPagercompletes. Previously, processing of pages constantly lagged behind the retrieval.gobenchresults:Before and after: