Skip to content

docs(helm): clarify integrated deployment and operations - #574

Open
imbajin wants to merge 4 commits into
apache:masterfrom
hugegraph:docs/helm-vermeer-refresh-20261115
Open

imbajin wants to merge 4 commits into
apache:masterfrom
hugegraph:docs/helm-vermeer-refresh-20261115

Conversation

@imbajin

@imbajin imbajin commented Nov 16, 2026

Copy link
Copy Markdown
Member

Purpose of the PR

Update the bilingual Helm deployment and operations guides for the integrated Hubble, Vermeer and optional GraphRAG chart. The guides now explain persistent bootstrap settings, credentials and recovery paths alongside the steps users execute.

  • Cover Vermeer discovery, task verification, matching Master/Worker versions and persistent role storage.
  • Explain RAG settings persistence and strict-mode network access for external models and browser traffic.
  • Make volume repair, maintenance and diagnostics easier to follow with native Steps, Tabs and Galleries.

Validation: strict build and independent artifact/link checks passed. The current chart ran in a minimal authenticated Kind deployment: five components Ready, five PVCs Bound, installation and test hooks passed, and five Hubble-to-Vermeer algorithms completed with paged results. Chart rendering/lint/unit checks and desktop/mobile interactions passed.

The single-node run does not establish HA or recovery acceptance. RAG configuration was rendered without connecting a live model. Runtime image identities and six shared screenshots are recorded in the fixture.

Dependent guides are separated into Computing #71 and AI #74, both based on this branch. Merge this deployment foundation before retargeting those guides to master.

@bitflicker64 bitflicker64 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Blocking: no. Summary: Vermeer, RAG, bootstrap-record, Text2GQL key, Store lifecycle and Vermeer version claims match the chart, Hubble, Store and Vermeer sources on master, and the install commands render as described. Three small accuracy and maintainability gaps remain.

Comment thread scripts/helm-refresh-evidence.md Resolved

@bitflicker64 bitflicker64 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Blocking: no. Score: 9/10. Summary: Vermeer, RAG, bootstrap-record, Text2GQL key, Store lifecycle and NetworkPolicy claims match the chart, Hubble, Store and Vermeer sources on master, and both topology installs render as written. The storeReplicas, Test saved connection and evidence-path comments from the earlier head are resolved.

@imbajin imbajin left a comment

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Score: 7.0/10 · Confidence: medium

Blocking: yes

Summary

The volume repair instructions can restore the normal policy before OnDelete Store Pods have applied the recursive repair, leaving inaccessible files on their PVCs.

Evidence

The cluster preset sets Store to OnDelete; the new verification checks image names only, which can already match while the running Pod still uses the old security context.

Validation notes

Validated against the documentation and chart source; no live upgrade or recovery exercise was performed.

@bitflicker64 bitflicker64 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Blocking: no. Summary: The Vermeer, RAG, bootstrap-record, fsGroup repair, Text2GQL key, Store lifecycle and NetworkPolicy claims match the chart, Hubble, Store and Vermeer sources on master, all install commands render as written, and the four earlier comments are fixed at this head. One Chinese cross-page link points to a fragment that does not exist.


在 `values-cluster.yaml`(`store.updateStrategy.type=OnDelete`)预设下,`helm upgrade` 不会自动重建 Store Pod。
即使镜像名称匹配,旧 Pod 也并未以 `Always` 策略重新挂载 PVC。
请按[受控滚动升级流程](./hugegraph-helm-operations.md#6-安全逐个滚动更新-store-镜像)逐个删除并重建 Store Pod。

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Minor: the operations page has no #6-安全逐个滚动更新-store-镜像 anchor (section 6 renders as id=6-rolling-store-images-safely), so this link opens the top of the page instead of the Store rollout procedure. Change the fragment to #6-rolling-store-images-safely, as the English step does.

@imbajin imbajin left a comment

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Score: 8.2/10 · Confidence: high

Blocking: no

Summary

The existing Chinese Store rollout link issue remains: operators following volume repair cannot jump to the controlled replacement procedure. The previous OnDelete repair finding is addressed.

Evidence

The target heading declares a different explicit fragment; the current recovery steps now require one-at-a-time replacements and verification of Always and shard recovery.

Validation notes

Reviewed documentation and chart sources; did not exercise a live cluster upgrade, volume recovery, or browser navigation.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants