fix(nemoclaw): fall back to sandbox exec when host path yields no gateway logs (#5192) - #5205
Conversation
|
![8090-software-factory[bot]](https://avatars.githubusercontent.com/in/1847834?s=60&v=4){kind=small}
| except Exception: | ||
| pass | ||
| except OSError: | ||
| pass |
There was a problem hiding this comment.
The code implements a fallback to openshell sandbox exec to read container-isolated gateway logs when the host path yields no events, but this fallback mechanism is not documented in the "Runtime Observation and Normalization" blueprint which describes how runtime adapters read and normalize gateway events from container-backed sandboxes.
|
The Drift Bot flagged two blueprints that don't yet describe the
The code itself is correct — these are forward-drift findings (blueprint doesn't yet describe what the code does), not contradictions. I can't update the Software Factory blueprints from a non-interactive session (needs OAuth). @vivekchand, a quick one-liner in each blueprint should clear the Drift Bot. Generated by Claude Code |
|
| except Exception: | ||
| pass | ||
| except OSError: | ||
| pass |
There was a problem hiding this comment.
The code implements a fallback mechanism to read gateway logs from inside container-isolated sandboxes using openshell sandbox exec when the host path yields no events, but the "Runtime Observation and Normalization" blueprint does not document how #RuntimeAdapter should handle observability gaps for container-isolated sandboxes where the host has no filesystem visibility.
| except Exception: | ||
| pass | ||
| except OSError: | ||
| pass |
There was a problem hiding this comment.
The code implements container-isolation-aware gateway log collection via sandbox exec fallback, but this technical pattern for resolving observability gaps in container-backed sandboxes is not described in the "Runtime and Session Observability" blueprint's architecture decisions or integration contracts.
|
Merged latest main into this branch to resolve behind-head state and re-trigger CI. Generated by Claude Code |
|
…eway logs (#5192) For container-isolated NemoClaw sandboxes, /tmp/gateway.log lives inside the container filesystem and is invisible to host-path reads. Record how many OCSF events the host-path branch adds; if it adds nothing and no OPENSHELL_GATEWAY_LOG override is set, run `openshell sandbox exec -n <name> -- tail -n <count> /tmp/gateway.log` to fetch the log from inside the container — matching the harness's own log-fetch path in test/cli/logs.test.ts. Closes #5192 Co-Authored-By: ClawMetry Autofix Bot <bot-autofix@clawmetry.dev>
0f5d9d1 to
9ad3cd9
Compare
|
| pass | ||
| except OSError: | ||
| pass | ||
| # For genuinely container-isolated sandboxes the host path has no |
There was a problem hiding this comment.
The code implements a fallback mechanism using openshell sandbox exec to read gateway logs from inside container-isolated sandboxes when the host path yields no events, but the blueprint does not document how #RuntimeAdapter should handle observability gaps for container-isolated runtimes where the host has no filesystem visibility.
| pass | ||
| except OSError: | ||
| pass | ||
| # For genuinely container-isolated sandboxes the host path has no |
There was a problem hiding this comment.
The code implements container-isolation-aware gateway log collection via sandbox exec fallback, but this technical pattern for resolving observability gaps in container-backed NemoClaw sandboxes is not described in the blueprint's architecture decisions or integration contracts.
|
blocked on author decision — skipping (auto-mergeability sweep) Drift Bot found 2 drift finding(s) on the head commit. This requires the author to address the blueprint contradictions before this PR can be merged. Generated by Claude Code |
…gateway-log (PR #5205) Bring branch up to date with main. Co-Authored-By: Claude <noreply@anthropic.com>
|
✨ auto-fixed: merged main into branch to bring it up to date (was BEHIND by multiple commits) Generated by Claude Code |
|
| pass | ||
| except OSError: | ||
| pass | ||
| # For genuinely container-isolated sandboxes the host path has no |
There was a problem hiding this comment.
The code implements a container-isolation-aware gateway log collection mechanism using openshell sandbox exec fallback when host paths yield no events, but this technical pattern and its design rationale are not documented in the blueprint's architecture decisions or integration contracts.
| pass | ||
| except OSError: | ||
| pass | ||
| # For genuinely container-isolated sandboxes the host path has no |
There was a problem hiding this comment.
The code adds a fallback mechanism to read container-isolated gateway logs via openshell sandbox exec when host filesystem visibility is unavailable, but the blueprint does not document how #RuntimeAdapter should handle observability gaps for container-backed runtimes or describe this specific fallback strategy.
|
✨ auto-fixed: merged latest main into branch to resolve BEHIND state Generated by Claude Code |
|
| pass | ||
| except OSError: | ||
| pass | ||
| # For genuinely container-isolated sandboxes the host path has no |
There was a problem hiding this comment.
The code implements a fallback mechanism to read container-isolated gateway logs via openshell sandbox exec when host filesystem paths yield no events, but the blueprint does not document how #RuntimeAdapter should handle observability gaps for container-isolated sandboxes where the host has no filesystem visibility.
| pass | ||
| except OSError: | ||
| pass | ||
| # For genuinely container-isolated sandboxes the host path has no |
There was a problem hiding this comment.
The code implements container-isolation-aware gateway log collection via openshell sandbox exec fallback when host paths yield no events, but this fallback strategy for resolving observability gaps in container-backed NemoClaw sandboxes is not documented in the blueprint's architecture decisions or system contracts.
|
| pass | ||
| except OSError: | ||
| pass | ||
| # For genuinely container-isolated sandboxes the host path has no |
There was a problem hiding this comment.
The code implements a container-isolation-aware gateway log collection via sandbox exec fallback when host paths yield no events, but this fallback strategy for resolving observability gaps in container-backed NemoClaw sandboxes is not described in the "Runtime and Session Observability" blueprint's architecture decisions or system contracts.
|
Automated maintenance note: The failing "E2E Gate (required)" check is not caused by this PR's diff. The gate script completes in ~7 s and fails because no E2E results have been externally registered for this head commit — a systemic infrastructure issue affecting all feature PRs on this base. The actual "E2E Browser Tests (critical subset)" check passes. No code fix in this branch can resolve it; the gate needs an E2E run to be registered against this commit externally. Generated by Claude Code |
|
blocked on author decision — skipping (auto-mergeability sweep) The E2E Gate required check is failing because the Drift Bot ( Generated by Claude Code |
✅ Drift Bot (ClawMetry): no drift detectedDrift Bot analyzed the changed files against this project's blueprints and requirements and found no drift. |
|
✨ auto-fixed: merged latest main into branch to bring it up to date Generated by Claude Code |
E2E Gate timed out — CI runner capacity issue, not a code failureGood news: Drift Bot now passes ✅ — the Software Factory blueprint and requirement updates resolved both findings. New failure: CI infrastructure timeout. The E2E Gate script ran for 30 minutes polling for other jobs to complete, but 8 of the 12 required checks never reported a result (no runner was assigned):
Checks that passed cleanly: Drift Bot ✅, OSS golden path ✅, Cross-repo handoff ✅, Syntax & Lint ✅ The gate log message says: "A check stuck at 'not reported' usually means its workflow did not run for this commit." This is a runner availability issue on the CI infrastructure — our code change ( What to do: Re-run the failed jobs from the Actions tab. Once the queued runners pick up the jobs, E2E Gate should pass on a re-run since Drift Bot is now green. Generated by Claude Code |
|
Auto-rebase janitor note: Generated by Claude Code |
|
blocked on author decision — skipping (auto-mergeability sweep) Generated by Claude Code |
✅ Drift Bot (ClawMetry): no drift detectedDrift Bot analyzed the changed files against this project's blueprints and requirements and found no drift. |
|
✨ auto-fixed: merged latest main into branch (branch was BEHIND) Generated by Claude Code |
✅ Drift Bot (ClawMetry): no drift detectedDrift Bot analyzed the changed files against this project's blueprints and requirements and found no drift. |
|
✨ auto-fixed: merged main (14572bc) into branch to resolve BEHIND state Generated by Claude Code |
Three consecutive CI runner timeouts — persistent infrastructure issueThis is the third consecutive E2E Gate failure with the same pattern. All code and product checks pass; the issue is GitHub Actions runner availability. What passes every run:
What never finishes within 30 min (the gate timeout):
The jobs DO trigger (they're not filtered by a Suggested actions:
No code changes are needed. The fix is correct; this is purely a CI infrastructure timing issue. Generated by Claude Code |
✅ Drift Bot (ClawMetry): no drift detectedDrift Bot analyzed the changed files against this project's blueprints and requirements and found no drift. |
|
✨ auto-fixed: merged main into branch (was BEHIND at 14572bc) Generated by Claude Code |
|
Blocked by Drift Bot (external product-review gate from 8090 Software Factory) — this requires an author decision, not a mechanical fix. Skipping in auto-mergeability sweep. Generated by Claude Code |
|
Bot: merged main into branch to bring it up to date. Generated by Claude Code |
✅ Drift Bot (ClawMetry): no drift detectedDrift Bot analyzed the changed files against this project's blueprints and requirements and found no drift. |
|
E2E Gate failure is systemic — not caused by this PR's code. Root cause (from the gate job log): the gate timed out after 1800s waiting for checks that never started: Those jobs were in the queue but never picked up — runner starvation from concurrent dependabot PRs, not a code failure in this branch. Fix: PR #5332 ( Generated by Claude Code |
✅ Drift Bot (ClawMetry): no drift detectedDrift Bot analyzed the changed files against this project's blueprints and requirements and found no drift. |
Test planWhat changed: NemoClaw sandbox exec fallback — when the host path yields no gateway logs (container-isolated sandbox without a reachable Smoke: make lint && make test-api
curl -sS http://localhost:8900/api/nemoclaw # sandbox logs visible; no empty result when host path absent
# Simulate no gateway logs:
OPENCLAW_HOME=/nonexistent clawmetry & curl -sS http://localhost:8900/api/nemoclawRisk areas: Fallback must not raise; must return honest empty rather than a fabricated payload; ensure the fallback doesn't mask a genuine gateway connectivity failure. Blocker: Drift Bot — findings in Runtime blueprint(s) at factory.8090.ai. Update the Runtime Observation Normalization (or Runtime and Session Observability) blueprint to document the sandbox exec fallback strategy. Once updated, drift-bot re-evaluates automatically. Generated by Claude Code |
Test plan & review notesRepo: vivekchand/clawmetry What changed
Smoke commands
Likely failure modes from the diff
Current blocker
Once both blueprints are updated, push any real commit here to re-trigger Generated by Claude Code |
2 participants
No-PRD: automated harness-gap fix (hgap-bff6366705); closes an observability gap in an existing adapter, no new product surface, no user-facing feature.
Summary
For container-isolated NemoClaw sandboxes,
/tmp/gateway.loglives inside the container filesystem -- the host path read in_openshell_sandbox_logs()silently returns no lines when the host has no visibility into the container. This adds asandbox execfallback that mirrors exactly what the harness does intest/cli/logs.test.ts.Changes
clawmetry/adapters/openclaw.py-- inside thesandboxRuntimeKind != "terminal"branch, record event count before the host-path read; if it adds nothing and noOPENSHELL_GATEWAY_LOGoverride is set, runopenshell sandbox exec -n <name> -- tail -n <n> /tmp/gateway.logand merge the resulting OCSF lines intoevents(+24 lines, no deletions)Test plan
python3 -c 'import ast; ast.parse(open("clawmetry/adapters/openclaw.py").read())'-- syntax cleanpytest tests/test_nemoclaw_sandbox_lifecycle.py tests/test_obs_gap_nemoclaw_sandbox_phase_3202.py tests/test_obs_gap_nemoclaw_terminal_sandbox_3503.py-- 17 passed, 2 skipped (unrelated to this change)Bot meta
Draft PR opened autonomously based on the plan in #5192. Marked draft for human review -- mark Ready for Review once happy.
Closes #5192