[ISSUE #10441] Reduce per-RPC allocation in metrics by caching static AttributeKey instances

[wang-jiahua]

Cache OpenTelemetry AttributeKey instances as static finals instead of creating new instances per RPC call. Also optimize TopicMessageType lookup, add volatile inline cache for RemotingCodeDistributionHandler, and add Logback NopStatusListener to suppress startup log noise.

Which Issue(s) This PR Fixes

• Fixes [Enhancement] Reduce per-RPC allocation in metrics and remoting by caching static AttributeKey instances #10441

Brief Description

Cache OpenTelemetry AttributeKey instances as static finals instead of creating new ones per RPC call. JFR heap dump shows 38,182 AttributeKey instances with only 6 distinct key names — each RPC allocates 3-4 throwaway key objects.

Changes:

• BrokerMetricsConstant / RemotingMetricsConstant: String labels → AttributeKey<T> static finals
• BrokerMetricsManager / RemotingMetricsManager / PopMetricsManager: use put(AttributeKey, value) overloads
• RemotingCodeDistributionHandler: volatile inline cache for counter methods
• TopicMessageType: direct if-else lookup replacing values() iteration
• rmq.broker.logback.xml: add NopStatusListener

How Did You Test This Change?

• JFR profiling (ObjectAllocationSample) on 4×ECS (8c/30G), 128 producer threads, 1KB body, Temurin JDK 21
• Verified all existing metrics report identically before/after
• Zero functional change — only metric label construction paths affected

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR focuses on reducing allocation/CPU overhead on metrics and remoting hot paths by introducing AttributeKey singletons and adding small caches for frequently repeated attribute/metric lookups.

Changes:

• Added fast-path caching for remoting request/response code distribution counting.
• Introduced typed OpenTelemetry AttributeKey constants and attribute caching/build helpers in remoting and broker metrics.
• Reduced repeated string allocations (e.g., cached lowercase metrics value, avoided Map.getOrDefault boxing paths).

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
remoting/src/main/java/org/apache/rocketmq/remoting/netty/RemotingCodeDistributionHandler.java	Adds “last code” fast path to reduce map lookups for hot request/response codes.
remoting/src/main/java/org/apache/rocketmq/remoting/metrics/RemotingMetricsManager.java	Adds attribute caching and typed label keys to reduce OpenTelemetry attribute overhead.
remoting/src/main/java/org/apache/rocketmq/remoting/metrics/RemotingMetricsConstant.java	Introduces typed AttributeKey singletons for labels.
remoting/src/main/java/org/apache/rocketmq/remoting/common/RemotingHelper.java	Minor optimization to avoid getOrDefault overhead for code descriptions.
common/src/main/java/org/apache/rocketmq/common/attribute/TopicMessageType.java	Caches lowercase metrics value to avoid repeated conversions.
broker/src/main/resources/rmq.broker.logback.xml	Suppresses Logback status output via NopStatusListener.
broker/src/main/java/org/apache/rocketmq/broker/metrics/PopMetricsManager.java	Switches to typed label keys for attributes.
broker/src/main/java/org/apache/rocketmq/broker/metrics/BrokerMetricsManager.java	Adds topic attributes cache and switches to typed label keys.
broker/src/main/java/org/apache/rocketmq/broker/metrics/BrokerMetricsConstant.java	Introduces typed AttributeKey singletons for broker metrics labels.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 28.82883% with 79 lines in your changes missing coverage. Please review.
✅ Project coverage is 47.95%. Comparing base (82a6a78) to head (52c49e8).

Files with missing lines	Patch %	Lines
.../rocketmq/broker/metrics/BrokerMetricsManager.java	4.76%	40 Missing ⚠️
...ketmq/remoting/metrics/RemotingMetricsManager.java	0.00%	24 Missing ⚠️
...etmq/remoting/metrics/RemotingMetricsConstant.java	0.00%	5 Missing ⚠️
...che/rocketmq/broker/metrics/PopMetricsManager.java	33.33%	4 Missing ⚠️
...pache/rocketmq/remoting/common/RemotingHelper.java	0.00%	4 Missing ⚠️
...emoting/netty/RemotingCodeDistributionHandler.java	85.71%	0 Missing and 2 partials ⚠️

Additional details and impacted files

@@              Coverage Diff              @@
##             develop   #10443      +/-   ##
=============================================
- Coverage      48.05%   47.95%   -0.10%     
+ Complexity     13311    13288      -23     
=============================================
  Files           1377     1377              
  Lines         100632   100707      +75     
  Branches       12995    13006      +11     
=============================================
- Hits           48359    48295      -64     
- Misses         46351    46467     +116     
- Partials        5922     5945      +23     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[oss-sentinel-ai]

Review by github-manager-bot

Summary

This PR caches OpenTelemetry AttributeKey instances as static finals to avoid per-RPC allocation, optimizes TopicMessageType lookup, adds a volatile inline cache for RemotingCodeDistributionHandler, and adds Logback NopStatusListener to suppress startup log noise.

Findings

• [Good] BrokerMetricsConstant.java:69-86 — Pre-built typed AttributeKey singletons eliminate per-call allocation on the metrics hot path.
• [Good] BrokerMetricsManager.java:105-120 — Import changes correctly reference the new cached keys.
• [Good] RemotingCodeDistributionHandler.java:35-68 — Volatile inline cache pattern correctly avoids ConcurrentHashMap.computeIfAbsent() on repeated request codes.
• [Good] TopicMessageType.java — Enum-based lookup is more efficient than string comparison.
• [Good] rmq.broker.logback.xml — NopStatusListener suppresses Logback startup noise in production.

Suggestions

1. The volatile inline cache pattern in RemotingCodeDistributionHandler is effective for sequential workloads. Consider documenting that this optimization assumes temporal locality in request codes.
2. Consider adding a microbenchmark to quantify the allocation reduction.

Verdict

✅ Approved — Solid performance optimization with clean implementation.

---

Automated review by github-manager-bot