FORGE-WP-0011: Forgejo and runner resource requests (CUST-WP-0071, RESOURCE-WP-0008)

Receive the wave-1 allocation review rows: BEFORE evidence captured read-only, reviewed patch against the live railiance-apps deploy path, deploy/verify/ rollback commands. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Assistant: claude-code Assistant-Model: opus Assistant-Process: 8340@bnt-lap001 Assistant-Session: 2e192b3d-b3e8-4403-94d4-d55a2c8b476c
2026-10-21 08:50:42 +02:00 · 2026-10-21 08:50:42 +02:00 · 6d795677b8
commit 6d795677b8
parent 256da9fb70
12 changed files with 504 additions and 0 deletions
--- a/docs/changes/FORGE-WP-0011/railiance-apps-forgejo-resources.patch
+++ b/docs/changes/FORGE-WP-0011/railiance-apps-forgejo-resources.patch
@ -0,0 +1,59 @@
+diff --git a/helm/forgejo-values.yaml b/helm/forgejo-values.yaml
+index 3bb9812..08a219e 100644
+--- a/helm/forgejo-values.yaml
+++ b/helm/forgejo-values.yaml
+@@ -4,6 +4,19 @@
+ strategy:
+   type: Recreate
+ 
+# Main Forgejo container sizing (FORGE-WP-0011, CUST-WP-0071, RESOURCE-WP-0008).
+# Evidence 2026-10-21 (7d): CPU p95 133-160m, peak ~1030m; memory working set
+# p95 ~956Mi, per-pod peak 1252Mi, RSS peak 500Mi; no OOM events.
+# Request = p95 rule (200m / 1104Mi). Memory limit 1536Mi = 1.23x working-set
+# peak and ~3x RSS peak (the rest is reclaimable page cache). No CPU limit:
+# bursts to ~1 core are normal for git pack/clone work and must not throttle.
+resources:
+  requests:
+    cpu: 200m
+    memory: 1104Mi
+  limits:
+    memory: 1536Mi
+
+ service:
+   http:
+     type: ClusterIP
+diff --git a/manifests/forgejo-runner.yaml b/manifests/forgejo-runner.yaml
+index ccbb94b..a886adb 100644
+--- a/manifests/forgejo-runner.yaml
+++ b/manifests/forgejo-runner.yaml
+@@ -100,6 +100,16 @@ spec:
+             - -H
+             - tcp://127.0.0.1:2375
+             - --tls=false
+          # FORGE-WP-0011: job workload runs here (one job at a time,
+          # runner capacity: 1). Request covers idle p95 (391Mi); the limit
+          # bounds a build burst. Peak 7d: 200m / 693Mi working set.
+          resources:
+            requests:
+              cpu: 100m
+              memory: 432Mi
+            limits:
+              cpu: 1500m
+              memory: 2Gi
+           readinessProbe:
+             exec:
+               command: ["sh", "-c", "wget -q -O- -T 3 http://127.0.0.1:2375/_ping | grep -qx OK"]
+@@ -108,6 +118,13 @@ spec:
+         - name: runner
+           image: code.forgejo.org/forgejo/runner@sha256:5071e6832313bafe71577e05631bece88caff08fcfb372193e4a21941f4ed54b
+           imagePullPolicy: IfNotPresent
+          # FORGE-WP-0011: runner daemon; p95 2m / 56Mi, peak 124m / 459Mi
+          # (host-label jobs run in this container). Pod request total
+          # 125m / 496Mi. No limit set in this change (see FORGE-WP-0011).
+          resources:
+            requests:
+              cpu: 25m
+              memory: 64Mi
+           env:
+             - name: DOCKER_HOST
+               value: tcp://127.0.0.1:2375