Sienna Meridian Satterwhite
2e89854f86
Adds the impress Helm chart (suitenumerique/docs, v4.5.0) to the lasuite
namespace with full Pingora routing, VSO secrets, and local overlay
resource tuning.
Routing (pingora-config.yaml):
- docs.* frontend -> docs-frontend:80 (nginx, static Next.js export)
- /api/* and /admin/* -> docs-backend:80 (Django/uvicorn)
- /collaboration/ws/* -> docs-y-provider:4444 (Hocuspocus WebSocket)
- integration.* -> integration:80 (La Gaufre hub, same file)
Secrets (vault-secrets.yaml):
- VaultDynamicSecret docs-db-credentials (DB engine, static role)
- VaultStaticSecret docs-django-secret (DJANGO_SECRET_KEY)
- VaultStaticSecret docs-collaboration-secret (y-provider shared secret)
OIDC client (oidc-clients.yaml):
- Fix redirect_uri from /oidc/callback/ to /api/v1.0/callback/ -- impress
mounts all OIDC URLs under api/{API_VERSION}/ via lasuite.oidc_login,
same pattern as people.
Local overlay (values-resources.yaml):
- docs-backend: 512Mi limit, WEB_CONCURRENCY=2 (4 uvicorn workers
exceeded 384Mi at startup on the arm64 Lima VM)
- docs-celery-worker: 384Mi limit, CELERY_WORKER_CONCURRENCY=2
- docs-y-provider: 256Mi limit
- seaweedfs-filer: raised from 256Mi to 512Mi (OOMKilled during 188MB
multipart S3 upload of impress-y-provider image layer)
Local overlay (kustomization.yaml):
- Image mirrors for impress-backend, impress-frontend, impress-y-provider
(amd64-only images retagged to Gitea via cmd_mirror before deploy)
258 lines
4.6 KiB
YAML
258 lines
4.6 KiB
YAML
# Patch: apply §10.7 memory limits to all Deployments in the local overlay.
|
|
# These are intentionally tight to stay within the 12 GB Lima VM budget.
|
|
#
|
|
# Applied as a strategic merge patch. Each stanza targets one Deployment by name.
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: cloudnative-pg
|
|
namespace: data
|
|
spec:
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: manager
|
|
resources:
|
|
limits:
|
|
memory: 256Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: livekit-server
|
|
namespace: media
|
|
spec:
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: livekit-server
|
|
resources:
|
|
limits:
|
|
memory: 128Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: pingora
|
|
namespace: ingress
|
|
spec:
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: pingora
|
|
resources:
|
|
limits:
|
|
memory: 128Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: valkey
|
|
namespace: data
|
|
spec:
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: valkey
|
|
resources:
|
|
limits:
|
|
memory: 64Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: opensearch
|
|
namespace: data
|
|
spec:
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: opensearch
|
|
env:
|
|
# Reduce JVM heap so it fits within the 512Mi container limit.
|
|
# Base sets -Xms512m -Xmx1g which immediately OOMs the container.
|
|
- name: OPENSEARCH_JAVA_OPTS
|
|
value: "-Xms192m -Xmx256m"
|
|
resources:
|
|
limits:
|
|
memory: 512Mi
|
|
requests:
|
|
memory: 256Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: seaweedfs-filer
|
|
namespace: storage
|
|
spec:
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: filer
|
|
resources:
|
|
limits:
|
|
memory: 512Mi
|
|
requests:
|
|
memory: 128Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: hydra-hydra-maester
|
|
namespace: ory
|
|
spec:
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: hydra-maester
|
|
resources:
|
|
limits:
|
|
memory: 64Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: login-ui
|
|
namespace: ory
|
|
spec:
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: login-ui
|
|
resources:
|
|
limits:
|
|
memory: 192Mi
|
|
requests:
|
|
memory: 64Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: hive
|
|
namespace: lasuite
|
|
spec:
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: hive
|
|
resources:
|
|
limits:
|
|
memory: 64Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: people-backend
|
|
namespace: lasuite
|
|
spec:
|
|
replicas: 1
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: people-celery-worker
|
|
namespace: lasuite
|
|
spec:
|
|
replicas: 1
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: people-frontend
|
|
namespace: lasuite
|
|
spec:
|
|
replicas: 1
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: docs-celery-worker
|
|
namespace: lasuite
|
|
spec:
|
|
replicas: 1
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: docs
|
|
env:
|
|
# Celery workers: 2 concurrent workers fits within local memory budget.
|
|
- name: CELERY_WORKER_CONCURRENCY
|
|
value: "2"
|
|
resources:
|
|
limits:
|
|
memory: 384Mi
|
|
requests:
|
|
memory: 128Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: docs-backend
|
|
namespace: lasuite
|
|
spec:
|
|
replicas: 1
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: docs
|
|
env:
|
|
# 2 uvicorn workers instead of the default 4 to stay within local memory budget.
|
|
# Each worker loads the full Django+impress app (~150 MB), so 4 workers
|
|
# pushed peak RSS above 384 Mi and triggered OOMKill at startup.
|
|
- name: WEB_CONCURRENCY
|
|
value: "2"
|
|
resources:
|
|
limits:
|
|
memory: 512Mi
|
|
requests:
|
|
memory: 192Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: docs-frontend
|
|
namespace: lasuite
|
|
spec:
|
|
replicas: 1
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: docs
|
|
resources:
|
|
limits:
|
|
memory: 128Mi
|
|
|
|
---
|
|
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: docs-y-provider
|
|
namespace: lasuite
|
|
spec:
|
|
replicas: 1
|
|
template:
|
|
spec:
|
|
containers:
|
|
- name: docs
|
|
resources:
|
|
limits:
|
|
memory: 256Mi
|
|
requests:
|
|
memory: 64Mi
|