Skip to content

docs(inbox): mini proc-exhaustion → ssh channel-reject (pool host-health guard 권고)#2

Open
dancinlife wants to merge 1 commit into
mainfrom
inbox/mini-proc-exhaustion-ssh-channel-reject-2026-05-24
Open

docs(inbox): mini proc-exhaustion → ssh channel-reject (pool host-health guard 권고)#2
dancinlife wants to merge 1 commit into
mainfrom
inbox/mini-proc-exhaustion-ssh-channel-reject-2026-05-24

Conversation

@dancinlife
Copy link
Copy Markdown
Contributor

Summary

  • pool roster host mini (macos) 에서 ssh mini '<cmd>' / pool on mini <cmd> 가 burst 시 간헐적으로 exec request failed on channel 0 실패.
  • Root cause (read-only 진단): uid mini proc-exhaustion — 2380 leaked sh (2616/2666 kern.maxprocperuid). 누수원은 anima akida_bridge.bin daemon 의 subshell spawn-without-reap (별도 anima-side fix).
  • POOL-side 관심사: pool 에 host-health guard 가 없어 슬롯 고갈된 host 에 계속 dispatch 했고, channel-reject 에러가 불투명. inbox 패치로 3-candidate resilience fix 제안.

3-candidate fix (g0 우선순위)

  • (a) host-health preflight — dispatch 전 ssh <host> 'ps -u $USER | wc -l' vs kern.maxprocperuid 체크, >90% full 시 warn/skip (최우선).
  • (b) channel-reject 진단 표면화exec request failed on channel 0 catch 시 "host may be at proc-uid limit" hint emit.
  • (c) retry-with-backoff — transient reject → 짧은 delay 후 1회 retry.

Note

  • review-only inbox 패치 (commons @d g54). 근저 누수는 anima-side (akida_bridge.hexa subshell cleanup) 로 별도 filed — 이 PR 은 pool resilience 한정.
  • Cross-link: anima dancinlab/anima#203 · 선행 inbox/troubleshooting/2026-05-22-ubu-1-publickey-denied.md (pool health verb 후보의 second 실증 case).

🤖 Generated with Claude Code

…lth guard 권고)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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