add rss pressure cache controller - #24665
Merged
mergify[bot] merged 3 commits intoMay 28, 2026
Merged
Conversation
LeftHandCold
requested review from
XuPeng-SH,
fengttt and
gouhongshen
as code owners
May 28, 2026 07:30
Qodo reviews are paused for this user.Troubleshooting steps vary by plan Learn more → On a Teams plan? Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center? |
There was a problem hiding this comment.
Pull request overview
This PR adds an explicit RSS pressure controller for memory cache admission and eviction, replacing TTL-only lifecycle behavior with state-based pressure handling.
Changes:
- Adds RSS pressure states (
none,soft,hard) with hysteresis thresholds and target percentages. - Splits pressure target set/clear from cache eviction and adds generation checks for async eviction.
- Exports memory cache pressure target clearing and wires the new pressure target callbacks in CN and workspace throttlers.
Reviewed changes
Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
pkg/common/rscthrottler/resource_throttler.go |
Implements RSS pressure state machine, target lifecycle callbacks, and async evict generation guard. |
pkg/common/rscthrottler/resource_throttler_test.go |
Adds lifecycle coverage for state transitions and updates concurrent escalation test inputs. |
pkg/fileservice/mem_cache.go |
Exports pressure target clearing and reuses it for invalid target inputs. |
pkg/fileservice/mem_cache_test.go |
Adds coverage for clearing memory cache pressure targets. |
pkg/cnservice/server.go |
Wires CN RSS cache pressure target setter/clearer separately from eviction. |
pkg/vm/engine/disttae/engine.go |
Wires workspace RSS cache pressure target setter/clearer separately from eviction. |
Contributor
Author
|
Addressed the review findings:
Re-ran focused tests: GOCACHE=/private/tmp/mo-gocache-step3 go test -mod=mod ./pkg/common/rscthrottler -run 'TestMemThrottlerRSS' -count=1
GOCACHE=/private/tmp/mo-gocache-step3 go test -mod=mod ./pkg/fileservice -run 'TestMemoryCachePressure|TestClearMemoryCachePressureTarget|TestMemoryCacheEvictToCapacityPercent' -count=1
GOCACHE=/private/tmp/mo-gocache-step3 go test -mod=mod ./pkg/cnservice -run 'TestMakeRSSCacheEvictorEvictsMemoryCacheOnly' -count=1
GOCACHE=/private/tmp/mo-gocache-step3 go test -mod=mod ./pkg/vm/engine/disttae -run 'TestNonExistent' -count=1 |
LeftHandCold
changed the title
gouhongshen
approved these changes
May 28, 2026
gouhongshen
left a comment
gouhongshen
left a comment
Contributor
There was a problem hiding this comment.
Code-review skill result (full review mode): no must-fix / merge-blocking issues found.\n\nDestructive-test conclusion: PARTIAL. The PR adds focused lifecycle and owner-isolation tests, but does not include destructive stress tests for concurrent owner target churn or deadline/ownership races under continuous pressure transitions (e.g., interleaving SetByOwner and ClearByOwner under concurrent scavenge calls).\n\nNon-blocking findings:\n1) Consider adding a concurrency-focused destructive test around owner-target interleaving to validate effective target recomputation under high-frequency transitions.\n2) Consider adding a test for expired-owner cleanup during repeated pressure churn to confirm effective deadline handling remains correct under continuous updates.\n\nThese findings are non-blocking and do not prevent merge.
XuPeng-SH
approved these changes
May 28, 2026
Contributor
Merge Queue Status
This pull request spent 1 minute 9 seconds in the queue, including 13 seconds running CI. Required conditions to merge
|
XuPeng-SH
added a commit
that referenced
this pull request
May 28, 2026
### What type of PR is this? /kind improvement ### What this PR does / why we need it: Cherry-pick #24665 to 4.0-dev. This backports the RSS pressure controller state machine and cache pressure target lifecycle improvements: - use none / soft / hard RSS pressure states with hysteresis; - keep memory cache pressure target as a lease while RSS pressure is active, and clear it immediately after pressure exits; - separate action throttling from pressure state transitions; - use owner-scoped memory cache pressure targets so CN and Workspace controllers do not clear each other; - keep async cache eviction guarded by pressure generation to avoid stale eviction actions from rewriting controller state. ### Which issue(s) this PR fixes: Refs #24664 Backport of #24665 ### Special notes for your reviewer: The cherry-pick had one conflict in pkg/common/rscthrottler/resource_throttler.go. The conflict was resolved by keeping the main-branch generation guard for async cache eviction. ### Does this PR introduce a user-facing change? NONE ### How to test: - GOCACHE=/private/tmp/mo-gocache-pr24665-4dev go test ./pkg/common/rscthrottler ./pkg/fileservice ./pkg/cnservice -run 'TestMemThrottlerRSS|TestMemoryCachePressure|TestClearMemoryCachePressureTarget|TestMakeRSSCacheEvictor' -count=1 Co-authored-by: XuPeng-SH <xupeng3112@163.com>
11 tasks
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
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.
What type of PR is this?
Which issue(s) this PR fixes:
Fixes #24664
What this PR does / why we need it:
This PR adds an RSS pressure controller for memory cache pressure handling.
Before this change, RSS cache pressure used a fixed admission-target TTL as the main lifecycle mechanism. If RSS recovered quickly, the memory cache admission target could still remain active until the TTL expired, slowing data cache recovery.
This PR changes that behavior by:
none,soft, andhard;none -> soft: RSS >= 85%;soft -> hard: RSS >= 92%;hard -> soft: RSS <= 88%;soft/hard -> none: RSS <= 80%;debug.FreeOSMemoryactions throttled by cooldown;The change keeps Step 1/2 behavior intact: RSS pressure still only controls data/memory cache and does not force objectio meta cache eviction.