Skip to content

fix(admins-openclaw): revert to beta.1 — beta.2 has HTTP handler regression - #60

Merged
mashkovd merged 1 commit into
mainfrom
fix/admins-openclaw-revert-to-beta.1
Apr 28, 2026
Merged

mashkovd merged 1 commit into
mainfrom
fix/admins-openclaw-revert-to-beta.1

Conversation

@mashkovd

Copy link
Copy Markdown
Member

Summary

  • Per chore(admins-openclaw): roll forward to 2026.4.28-beta.2 #57 + fix(admins-openclaw): bump CPU limit and startup probe timeout for beta.2 #59 we tried rolling admins-openclaw to `2026.4.28-beta.2`. After fixing probe timeouts and bumping CPU limit to the tenant max (2000m), the gateway STILL fails to serve /healthz:
    • TCP listener on `:18789` is up — `curl -v` connects successfully
    • GET /healthz request is sent, then server never responds (10s timeout from inside the pod)
    • CPU usage sits at 1.3-1.7 cores out of 2.0 — NOT throttled
  • This is a real HTTP-handler regression in the gateway code shipped between beta.1 and beta.2, not a config issue. No probe/CPU tuning will fix it.
  • Reverting admins to `2026.4.26-beta.1` (matching labs/ovk).

Diagnostic evidence

```
$ kubectl exec base-service -- curl -v -m 10 http://127.0.0.1:18789/healthz

  • Trying 127.0.0.1:18789...
  • Connected to 127.0.0.1 (127.0.0.1) port 18789 (#0)

GET /healthz HTTP/1.1
Host: 127.0.0.1:18789
...

  • Operation timed out after 10000 milliseconds with 0 bytes received

$ kubectl top pod
admins-openclaw-base-service-... base-service 1581m ... (limit 2000m)
```

Changes

File Change
`platform-gitops/services/admins/openclaw/values.yaml` Image tag, `OPENCLAW_VERSION`, setup-init image: `2026.4.28-beta.2` → `2026.4.26-beta.1`

Probe tuning and CPU bump from #55/#59 stay in place — they're correct for beta.1 too and harmless headroom.

Test plan

  • After merge, ArgoCD syncs admins-openclaw
  • `kubectl -n admins get pod` → 3/3 Running, RestartCount=0 within 2 minutes
  • `curl -m 3 /healthz` returns 200 from inside pod
  • Soak 5 min — no restarts
  • `https://admins-openclaw.mctl.ai/\` → 302 to OIDC login

Follow-ups

  • File issue at `mctlhq/mctl-openclaw` describing the beta.2 HTTP-handler hang. Diff range `v2026.4.26-beta.1..v2026.4.28-beta.2` includes a major upstream sync (13k files) plus `fix(gateway): normalize MCP object schemas` and `fix(plugins): strip workspace:* devDeps` — start there.

…ndler regression

beta.2 admission test failed:
- TCP listener on :18789 binds and accepts connections
- After GET /healthz HTTP/1.1 is sent, server never responds (10s timeout)
- CPU sits at 1.3-1.7 cores out of 2-core limit (NOT throttled)

The HTTP request pipeline is wedged in beta.2 — no amount of probe/CPU
tuning fixes a non-responding /healthz handler. Rolling admins back to
beta.1 alongside labs/ovk while we file an issue against mctl-openclaw
to investigate the beta.2 regression.

Probe and CPU tuning from prior PRs stays in place — they're correct
fixes for beta.1's behavior too and harmless headroom for any future
build.
@mashkovd

Copy link
Copy Markdown
Member Author

@codex review

@chatgpt-codex-connector

Copy link
Copy Markdown

Codex Review: Didn't find any major issues. Can't wait for the next one!

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

@mashkovd
mashkovd merged commit 24da2e8 into main Apr 28, 2026
@mashkovd
mashkovd deleted the fix/admins-openclaw-revert-to-beta.1 branch April 28, 2026 21:56
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.

1 participant