Repository navigation
fix(hubble): clarify operations process metrics - #884
Conversation
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## master #884 +/- ##
============================================
+ Coverage 55.40% 55.54% +0.13%
+ Complexity 5989 3731 -2258
============================================
Files 619 285 -334
Lines 32582 19908 -12674
Branches 4217 3042 -1175
============================================
- Hits 18052 11057 -6995
+ Misses 12952 7671 -5281
+ Partials 1578 1180 -398 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
bitflicker64
left a comment
There was a problem hiding this comment.
Blocking: yes. Summary: The NodeDetail stale fix holds, but the Nodes page, where Overview links PD diagnostics, still shows a retained PD endpoint sample as current after a failed refresh.
bitflicker64
left a comment
There was a problem hiding this comment.
Blocking: no. Score: 9/10. Summary: Nodes and NodeDetail now mark retained sources stale after a failed refresh, which resolves both earlier PD endpoint sample threads, and the Store lifecycle probe stays bounded by the Store deadline with a 404 fallback for older Stores.
Before: losslessly parsed Vermeer CPU values could appear missing, a configured PD Service endpoint could not reliably identify the responding member, and Store process lifecycle was not shown separately from readiness.
After: Normalize finite nonnegative display measurements while preserving zero/unavailable states; show unmatched PD endpoint metrics as a bounded, separately labelled sample rather than a node or cluster total; collect optional Store process lifecycle independently of readiness with an old-version fallback.
The lifecycle endpoint requires a Store build containing 🔗 apache/hugegraph #3441; related documentation: 🔗 apache/hugegraph-doc #555. STARTED does not establish PD membership, partition health, or read/write readiness.
Source: 🔗 hugegraph/hugegraph-toolchain #127
Author-reported tests use controlled responses. Deployed PD Service, Vermeer and Store validation remains pending.
Visual overview
运维展示对有限非负度量值做边界转换并保留 0,不修改精确存储;未知或过期不冒充 0。无法唯一匹配成员时,PD Service 指标只作为有界、单独标注的端点样本,不归属某节点或计为集群总量。Store lifecycle 独立于原 readiness 采集,共用截止时间;STARTED 不证明 PD/分片/读写就绪,旧版 404 显示不支持。生命周期接口依赖 🔗 Server #3441,说明见 🔗 Doc #555。真实部署仍待验证,不表示 CI、独立现场或同一 RC 验收通过。 评审核对:2026-10-31。