Auto-sync: 2026-10-28
This commit is contained in:
+26
-7
@@ -40,12 +40,31 @@ LLM-Chokepoint und objektiver Telemetrie.
|
||||
Tag `fleet-v1` = **digest `sha256:32eda4136f62ef15920f9414c30dd33e9c7736b53a6f02c1658b5584c45a41b7`**
|
||||
(1,07 GB; Basis nousresearch/hermes-agent digest-gepinnt + jq; Dockerfile
|
||||
`clusters/main/fleet/agent-image/Dockerfile`; gebaut auf mgmt-runner .124).
|
||||
**Digest-Pin live** in OpenShell-Values (`sandbox.image.digest`, Commit 640e051,
|
||||
App-Spec + fleet-truth synchron) → gateway.toml `default_image` → ArgoCD
|
||||
Synced/Healthy. Pull-Secret `agent-fleet-registry-pull` out-of-band im NS fleet
|
||||
(default-SA imagePullSecrets; nicht im Git). E2E: Kubelet-Pull 34,6 s auf
|
||||
rke2-cp-01, Container-Start, jq 1.7. Push-Root-Causes + Runbook: siee
|
||||
`systems/gitea.md` „Container Registry".
|
||||
- **p1-image v5.1 (28.10., AKTUELL):** Tag `fleet-v5.1` = **digest
|
||||
`sha256:7f141019695cf38468c3930a0bf39523c09f0b0fad364364b1f7e4b74a23094d`**
|
||||
(Commits d60b7c0+55dcc9a+c0666eb). Fixes seit v5: (a) **NAR-Verfahren**
|
||||
(Owner-Mandat 28.10.): Netz-Freigaben über CISO-Antrag statt Scan-
|
||||
Improvisation — FLEET-POLICIES.md (ro im Image) + CISO-Triage-Kriterien
|
||||
+ Worker-Tool-Prompt-Verweis; `to: tower-owner` NUR durch ciso und nur
|
||||
`type: escalation|fyi`, kodiert in 3 Validierungsschichten (submit_task.py
|
||||
+ CM-Kopie fleet-submit-task, watcher_main.py, board-service server.py),
|
||||
E2E bewiesen (ciso→tower-owner durch, ceo geblockt); (b) **Worker
|
||||
Resume-on-Disconnect:** `Gateway disconnected` startet KEIN Re-Execute,
|
||||
sondern beobachtet die Session-DB bis 900 s und completed mit der finalen
|
||||
assistant-Nachricht — Turns laufen im Daemon weiter, Scan-Loops
|
||||
(RCA: 500m-Cgroup-Verhungern, IPC-Riss) sind der verbotene Auslöser;
|
||||
(c) reasoning_effort max + LLM_TIMEOUT 1800 s (aus v5 übernommen).
|
||||
Rollout 28.10. 19:0xZ: 9/9 recreate (Gateway-Restart vor Rollout —
|
||||
default_image wird erst bei Neustart gelesen), Digest je Pod verifiziert.
|
||||
**Digest-Pin live** in OpenShell-Values (`sandbox.image.digest`) +
|
||||
Prepuller-DS `clusters/main/fleet/agent-image-prepull.yaml` → gateway.toml
|
||||
`default_image` → ArgoCD Synced/Healthy. Board-Service 0.16 (Digest
|
||||
sha256:4ea9f7bf…, NAR-Lane + submit_task.py aus CM fleet-submit-task)
|
||||
+ Watcher 0.16 (Digest sha256:77657260…) synchron gepingt. Pull-Secret
|
||||
`agent-fleet-registry-pull` out-of-band im NS fleet (default-SA
|
||||
imagePullSecrets; nicht im Git). Push-Root-Causes + Runbook: siehe
|
||||
`systems/gitea.md` „Container Registry". SolDoc:
|
||||
`docs/solutions/bug-fixes/2026-10-28-fleet-gateway-disconnect-scan-cpu-nar.md`.
|
||||
- **OpenShell:** NS openshell, Pod openshell-0, Helm-Chart ghcr.io/nvidia 0.1.5
|
||||
digest-gepinnt; sandbox_image_pull_policy if_not_present; Workspace IMMER
|
||||
/sandbox, UID 1000. Helm-Pitfall: gatewayConfig-Mapping mit EINEM quoted dotted
|
||||
@@ -71,7 +90,7 @@ LLM-Chokepoint und objektiver Telemetrie.
|
||||
| chro | chro-v2 | glm-5-3-flash | fleet-vk-chro |
|
||||
| cmo | cmo-v2 | glm-5-3-flash | fleet-vk-cmo |
|
||||
| agile-coach | agile-coach | glm-5-3-flash | fleet-vk-agile-coach |
|
||||
| coo | coo-v2 | deepseek-v41-flash (reasoning_effort low) | fleet-vk-coo |
|
||||
| coo | coo-v2 | deepseek-v41-flash (reasoning_effort max) | fleet-vk-coo |
|
||||
| devops | devops-v2 | deepseek-v41-flash | fleet-vk-devops |
|
||||
| research | research | deepseek-v41-flash | fleet-vk-research |
|
||||
|
||||
|
||||
+1
-1
@@ -75,7 +75,7 @@ modified: "2026-09-27"
|
||||
- **Key:** `~/.ssh/id_ed25519_gitea-hermes-push` (hermes-agent@gitea-push)
|
||||
- **1Password:** "Gitea Hermes Push Key" (ID: lqaebdtewj5xatvpzfrkag73v4)
|
||||
- **Gitea Key ID:** 4 (title: hermes-agent-push, user: dominik)
|
||||
- **Verwendung:** `GIT_SSH_COMMAND="ssh -i ~/.ssh/id_ed25519_gitea-hermes-push -o StrictHostKeyChecking=no" git push ssh://git@10.0.30.202:22/dominik/iac-homelab.git HEAD:main`
|
||||
- **Verwendung:** `GIT_SSH_COMMAND="ssh -i ~/.ssh/id_ed25519_gitea-hermes-push -o StrictHostKeyChecking=no" git push ssh://git@10.0.30.200:22/dominik/iac-homelab.git HEAD:main` (SSH-LB-IP ist **10.0.30.200** — NICHT .202; .202 = alte/andere Adresse, Verbindung timed out. Beweis 28.10.: Port-22-Check von .124 und .213 OK auf .200, BLOCKED auf .202.)
|
||||
- **Wichtig:** Gitea HTTP (Port 3000) hat keine externe IngressRoute — SSH (Port 22 via LoadBalancer) ist der einzige zuverlässige Weg für Git-Pushes von außerhalb K8s.
|
||||
|
||||
## Known ArgoCD Issues (Sep 2026)
|
||||
|
||||
Reference in New Issue
Block a user