Route issue-core directly on railiance01
All checks were successful
CI Smoke / host-smoke (push) Successful in 0s
CI Smoke / container-smoke (push) Successful in 2s

This commit is contained in:
tegwick 2026-08-19 23:41:57 +02:00
parent 41a3fb8b81
commit 213e810cf8
3 changed files with 16 additions and 127 deletions

View file

@ -78,23 +78,19 @@ activity-core.
Railiance production uses `state-hub` unless an overlay explicitly sets
`rest` for an experiment.
### Known production failure (2026-07-21, reconfirmed 2026-07-21 T06) — rest path
### Production direct path (verified 2026-08-19)
`POST /issues/` returns **HTTP 503**:
Railiance uses the private cluster-local endpoint:
```text
Failed to connect to backend 'forgejo-inbox': Failed to connect to Gitea API
http://issue-core.issue-core.svc.cluster.local:8765
```
(Historical error string from the Forgejo-compatible client library — the
deployed forge is **Forgejo**, not Gitea.)
**Disposition (ACTIVITY-WP-0023-T06):** activity-core keeps global default
`state-hub` and does **not** flip production to `rest`. Path A is owned by
**issue-core**: rotate/fix `GITEA_BACKEND_TOKEN` (Forgejo backend PAT for the
forgejo-inbox connector — not the activity-core `ISSUE_CORE_API_KEY`
ingestion key). After issue-core proves `POST /issues/`**201**, operators may
opt in per definition / overlay only (WP-0022).
`rapp-issue-core` verification proved authenticated create **201**, close
**200**, dependency failure **503**, and recovery **201/200** at immutable image
digest `sha256:a56c80ccbd1690bc0bcea384165b1967704f0fdbaf1c506ce1082f56190b92cf`.
activity-core still keeps the global default `state-hub`; a healthy transport
does not change the opt-in policy for external Forgejo projection.
Internal scheduled automation should use **ops_run claim** (WP-0026), not wait
on this rest path.
@ -103,7 +99,7 @@ Smoke from worker (does not change sink env):
```bash
kubectl -n activity-core exec deploy/actcore-worker -- python -c '
# valid IssueCore body → expect 201 when path A is healthy, else 503 forgejo-inbox
# valid IssueCore body → expect 201; close the controlled issue after the check
'
```

View file

@ -14,9 +14,9 @@ data:
LLM_CONNECT_URL: http://llm-connect.activity-core.svc.cluster.local:8080
LLM_CONNECT_TIMEOUT_SECONDS: "300"
REPO_SCOPING_URL: http://repo-scoping.repo-scoping.svc.cluster.local:8020
ISSUE_CORE_URL: http://actcore-issue-core-bridge.activity-core.svc.cluster.local:8765
# ACTIVITY-WP-0021: state-hub until issue-core Forgejo PAT (GITEA_BACKEND_TOKEN)
# is rotated; switch back to "rest" after path A smoke returns 201.
ISSUE_CORE_URL: http://issue-core.issue-core.svc.cluster.local:8765
# External Forgejo projection remains opt-in even though the direct path is
# healthy. Internal automation continues to use State Hub by default.
ISSUE_SINK_TYPE: "state-hub"
# ACTIVITY-WP-0026 / ACT-ADR-005 — claimable ops_run on emit (not Forgejo)
OPS_RUN_QUEUE_ENABLED: "true"
@ -1030,117 +1030,6 @@ spec:
persistentVolumeClaim:
claimName: actcore-statehub-edge-outbox
---
apiVersion: v1
kind: Service
metadata:
name: actcore-issue-core-bridge
namespace: activity-core
labels:
app.kubernetes.io/name: actcore-issue-core-bridge
app.kubernetes.io/part-of: activity-core
spec:
selector:
app.kubernetes.io/name: actcore-issue-core-bridge
ports:
- name: http
port: 8765
targetPort: http
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: actcore-issue-core-bridge
namespace: activity-core
labels:
app.kubernetes.io/name: actcore-issue-core-bridge
app.kubernetes.io/part-of: activity-core
spec:
replicas: 1
selector:
matchLabels:
app.kubernetes.io/name: actcore-issue-core-bridge
template:
metadata:
labels:
app.kubernetes.io/name: actcore-issue-core-bridge
app.kubernetes.io/part-of: activity-core
spec:
hostNetwork: true
dnsPolicy: ClusterFirstWithHostNet
containers:
- name: proxy
image: activity-core:railiance01-prod
imagePullPolicy: Never
ports:
- name: http
containerPort: 18081
command:
- python
- -c
- |
from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
from urllib.error import HTTPError, URLError
from urllib.request import Request, urlopen
TARGET = "http://127.0.0.1:18765"
HOP_HEADERS = {"connection", "host", "keep-alive", "proxy-authenticate",
"proxy-authorization", "te", "trailers",
"transfer-encoding", "upgrade"}
class Proxy(BaseHTTPRequestHandler):
def do_GET(self):
self._proxy()
def do_POST(self):
self._proxy()
def do_PATCH(self):
self._proxy()
def _proxy(self):
length = int(self.headers.get("content-length", "0") or "0")
body = self.rfile.read(length) if length else None
headers = {
key: value
for key, value in self.headers.items()
if key.lower() not in HOP_HEADERS
}
request = Request(
TARGET + self.path,
data=body,
headers=headers,
method=self.command,
)
try:
timeout = 360 if self.command == "POST" else 30
with urlopen(request, timeout=timeout) as response:
payload = response.read()
self.send_response(response.status)
for key, value in response.headers.items():
if key.lower() not in HOP_HEADERS:
self.send_header(key, value)
self.end_headers()
self.wfile.write(payload)
except HTTPError as exc:
payload = exc.read()
self.send_response(exc.code)
self.end_headers()
self.wfile.write(payload)
except URLError as exc:
self.send_response(502)
self.end_headers()
self.wfile.write(str(exc).encode())
ThreadingHTTPServer(("0.0.0.0", 18081), Proxy).serve_forever()
readinessProbe:
httpGet:
path: /healthz
port: http
initialDelaySeconds: 5
periodSeconds: 10
timeoutSeconds: 5
failureThreshold: 6
---
---
apiVersion: batch/v1
kind: Job

View file

@ -33,6 +33,10 @@ issues). Set `rest` only for intentional issue-core projection when the backend
is healthy. `ISSUE_CORE_API_KEY` and `FORGEJO_TOKEN` are synced from OpenBao into
`actcore-runtime-secret` by ExternalSecrets.
`ISSUE_CORE_URL` uses the private cross-namespace Service directly:
`http://issue-core.issue-core.svc.cluster.local:8765`. The historical
host-network bridge and workstation tunnel are not part of this deployment.
**Ops run claim queue (ACTIVITY-WP-0026):** ConfigMap sets
`OPS_RUN_QUEUE_ENABLED=true`. After image + migrate job (alembic **0007**),
workers insert claimable `ops_runs` on emit. Full railiance checklist: