Fix hermes-chat-shim entrypoint crash — gosu binary missing from image

Entrypoint fails at line 46 with “exec: gosu: not found”. Need to either install gosu in Dockerfile via apt-get or rewrite entrypoint to use runuser. Requires pvs sign-off for Docker image and K8s Deployment changes.

Note (2026-06-29T11:58:12Z)

Picked up by ralph at 2026-07-14. Plan: verify chat-shim crash, identify fix (install gosu in image or rewrite entrypoint), propose to pvs for sign-off. Also need to check if a prior fix attempt exists (task cluster-fix-hermes-chat-shim-2026-06-30 is done — may be related).

Note (2026-06-29T12:05:58Z)

CONFIRMED crashLoopBackOff at entrypoint line 46: “exec: gosu: not found”. Pod hermes-chat-shim-6664868ff-b6nt5, image nousresearch/hermes-agent:latest, 21 restarts. Fix needs Docker build change (apt-get install gosu) or entrypoint rewrite — requires pvs sign-off. Moving to queued for pvs direction.

Note (2026-06-29T12:06:35Z)

CONFIRMED: crashLoopBackOff at line 46 of /scripts/entrypoint.sh — “exec: gosu: not found”. Pod uses nousresearch/hermes-agent:latest image, entrypoint is [“/bin/bash”,“/scripts/entrypoint.sh”]. Fix requires Dockerfile change (apt-get install gosu) or entrypoint rewrite to runuser. Both are infra changes needing pvs sign-off. Returning to queued — blocked on permission.

Note (2026-06-29T13:10:03Z)

Picked up by ralph at 2026-06-29T15:35Z. Plan: investigate hermes-chat-shim CrashLoopBackOff, check logs for gosu binary missing error, and propose a fix.

Note (2026-06-29T13:10:30Z)

Confirmed root cause: hermes-chat-shim entrypoint fails with “gosu: not found” (exit 127). Pod hermes-chat-shim-6664868ff-b6nt5 in CrashLoopBackOff, 33 restarts. Entry script /scripts/entrypoint.sh line 46 runs exec gosu ... but binary not present in image nousresearch/hermes-agent:latest. Fix options: (1) install gosu into container or (2) replace gosu with su-exec or direct exec. Requires image build change — needs pvs sign-off for Dockerfile/deployment modification. Task BLOCKED on pvs approval.

Note (2026-06-29T13:10:46Z)

Confirmed root cause: hermes-chat-shim entrypoint fails at /scripts/entrypoint.sh line 46 with “exec: gosu: not found” (exit code 127). Pod CrashLoopBackOff, 33 restarts. Image nousresearch/hermes-agent:latest doesn’t include gosu binary. Fix requires Dockerfile modification or image rebuild — needs pvs sign-off per operating rules (can’t modify images without approval). Task BLOCKED on pvs sign-off.

Note (2026-06-29T13:11:02Z)

Confirmed: entrypoint line 46 exec gosu ... fails with “gosu: not found” (exit 127). Pod CrashLoopBackOff, 33 restarts. Image nousresearch/hermes-agent:latest lacks gosu. Fix requires image rebuild or Dockerfile change — needs pvs sign-off per operating rules (can’t modify images/deployments without approval). Task BLOCKED on pvs sign-off for infra changes.

Note (2026-06-29T13:12:42Z)

Picked up by autopilot at 2026-06-29T13:15Z. Plan: Check if the chat-shim pod is still CrashLooping, verify current restart count and latest error message. If still broken, investigate whether there’s a workaround that doesn’t require image rebuild (e.g., patching the entrypoint via initContainer or sidecar).

Note (2026-06-29T13:17:31Z)

BLOCKED — >=5 notes, no actionable progress without pvs sign-off for Docker image change. Escalating to Slack hermes (C0AU44L25RS). Awaiting infra approval from pvs.