chrisfu
|
6f29deba99
|
Supabase/CNPG stability: pin Supabase to merlin, reduce CNPG to 2 instances, fix supabase_enabled
Changes address resource pressure (merlin OOM with 3 CNPG + Supabase + monitoring) and
the supabase_enabled=false override in silent mode that blocked Supabase deploy.
- conf/service/prole.cfg:
- SUPABASE_PV_NODE/SUPABASE_PRIMARY_NODE = merlin.prole.org (d002 storage)
- SUPABASE_STORAGE_CLASS = merlin-local-iscsi-d002
- CNPG_ELIGIBLE_NODES: removed pi.prole.org (pi NotReady, d003 inaccessible)
- k8s/prole/knoe-db.yaml:
- instances: 3 -> 2 (merlin + myrddin only, pi excluded)
- maxSyncReplicas: 1 -> 0 (safe for 2-node cluster)
- supabase/deploy.sh:
- default SUPABASE_PV_NODE to merlin.prole.org, base dir to /synology/d002/supabase
- add idempotent ALTER ROLE password sync after CREATE ROLE IF NOT EXISTS
- resolve live DB_PASSWORD from knoe-db-superuser secret before helm render
- supabase/helm/render_supabase.py:
- resolve SUPABASE_STORAGE_CLASS from env/cfg, apply to storage/analytics PVCs
- DATABASE_NAMESPACE resolution: check env + cfg Global section before discovery
- DB_PASSWORD from env takes priority over cfg (live cluster secret wins)
- default primary node = merlin.prole.org
- knoe/ui/screens/cluster.py: minor display update
Co-authored-by: Junie <junie@jetbrains.com>
|
2026-03-24 13:52:38 -07:00 |
|