From 3867916b8135fb1e66bda74eb221a4a7d7bd6983 Mon Sep 17 00:00:00 2001 From: mctl-deploy Date: Tue, 7 Jul 2026 00:35:31 +0200 Subject: [PATCH] fix: tighten openclaw and claude-remote memory limits to 2x request admins/labs/ovk openclaw and labs/claude-remote carried memory limits 2.5x-4x their requests, a ratio built up reactively during past CPU incidents rather than tuned for memory. Tenant quota usage data shows this headroom sitting unused. Trim to base-service's own 2x default ratio to reclaim quota-ceiling headroom for onboarding more tenants, and apply the same ratio in the openclaw onboarding template so new tenants don't inherit the old ratio. CPU limits are untouched. --- .../service-templates/openclaw/values.yaml.tpl | 5 ++++- platform-gitops/services/admins/openclaw/values.yaml | 5 ++++- platform-gitops/services/labs/claude-remote/values.yaml | 5 ++++- platform-gitops/services/labs/openclaw/values.yaml | 5 ++++- platform-gitops/services/ovk/openclaw/values.yaml | 5 ++++- 5 files changed, 20 insertions(+), 5 deletions(-) diff --git a/platform-gitops/argo-workflows/service-templates/openclaw/values.yaml.tpl b/platform-gitops/argo-workflows/service-templates/openclaw/values.yaml.tpl index a0fba9e3b..f9e4c7c76 100644 --- a/platform-gitops/argo-workflows/service-templates/openclaw/values.yaml.tpl +++ b/platform-gitops/argo-workflows/service-templates/openclaw/values.yaml.tpl @@ -24,7 +24,10 @@ resources: memory: 768Mi limits: cpu: 1500m - memory: 2560Mi + # 2x the request, matching base-service's default ratio — new tenants + # should not inherit the 3.33x ratio the existing openclaw instances + # carried before their 2026-07 memory-limit trim. + memory: 1536Mi # Recreate avoids quota deadlocks for single-pod tenant deployments. strategy: diff --git a/platform-gitops/services/admins/openclaw/values.yaml b/platform-gitops/services/admins/openclaw/values.yaml index ec42634e6..b75398b26 100644 --- a/platform-gitops/services/admins/openclaw/values.yaml +++ b/platform-gitops/services/admins/openclaw/values.yaml @@ -47,7 +47,10 @@ resources: memory: 768Mi limits: cpu: 4000m - memory: 2560Mi + # Memory limit was 3.33x the request; tenant quota usage showed this + # headroom unused. Trimmed to base-service's default 2x ratio to free + # quota ceiling for onboarding more tenants (2026-07). + memory: 1536Mi # Recreate avoids quota deadlocks for single-pod tenant deployments. strategy: type: Recreate diff --git a/platform-gitops/services/labs/claude-remote/values.yaml b/platform-gitops/services/labs/claude-remote/values.yaml index 089d1c882..c5440d338 100644 --- a/platform-gitops/services/labs/claude-remote/values.yaml +++ b/platform-gitops/services/labs/claude-remote/values.yaml @@ -19,7 +19,10 @@ resources: # PR-steward agent runs spike past 1000m and get CFS-throttled. Raised # 1000m→1500m (fits labs LimitRange max 3 and quota limits.cpu 12). cpu: 1500m - memory: 2Gi + # Memory limit was 4x the request (2Gi); tenant quota usage data showed + # this headroom sitting unused. Trimmed to base-service's default 2x + # ratio to free quota ceiling for onboarding more tenants (2026-07). + memory: 1024Mi podSecurityContext: fsGroup: 1000 runAsNonRoot: true diff --git a/platform-gitops/services/labs/openclaw/values.yaml b/platform-gitops/services/labs/openclaw/values.yaml index 3940ef7ea..5f7514926 100644 --- a/platform-gitops/services/labs/openclaw/values.yaml +++ b/platform-gitops/services/labs/openclaw/values.yaml @@ -43,7 +43,10 @@ resources: memory: 1024Mi limits: cpu: 3000m - memory: 2560Mi + # Memory limit was 2.5x the request; tenant quota usage showed this + # headroom unused. Trimmed to base-service's default 2x ratio to free + # quota ceiling for onboarding more tenants (2026-07). + memory: 2048Mi # Recreate avoids quota deadlocks for this tenant during rollout strategy: type: Recreate diff --git a/platform-gitops/services/ovk/openclaw/values.yaml b/platform-gitops/services/ovk/openclaw/values.yaml index c342028c9..2d9437b4d 100644 --- a/platform-gitops/services/ovk/openclaw/values.yaml +++ b/platform-gitops/services/ovk/openclaw/values.yaml @@ -32,7 +32,10 @@ resources: memory: 768Mi limits: cpu: 3000m - memory: 2560Mi + # Memory limit was 3.33x the request; tenant quota usage showed this + # headroom unused. Trimmed to base-service's default 2x ratio to free + # quota ceiling for onboarding more tenants (2026-07). + memory: 1536Mi # Recreate avoids quota deadlocks for single-pod tenant deployments. strategy: type: Recreate