fix(temporal): switch to PostgreSQL (CNPG ddb-cluster) instead of broken Cassandra/ES setup

k8s/applications/temporal/temporal-values.yaml

@@ -1,65 +1,28 @@
 # k8s/temporal/temporal-values.yaml
-# Temporal — workflow engine for story-crater backend async task orchestration.
+# Temporal — workflow engine  
-# Chart: temporal/temporal from https://go.temporal.io/helm-charts
+# Uses external CNPG PostgreSQL for persistence (ddb-cluster)
-#
+# Visibility via same PostgreSQL database
-# Uses Cassandra for default store (workflow history/events)
-# Uses Elasticsearch for visibility store (namespace/workflow queries)
-# This is the chart's native, well-tested configuration.
 
-# ── Datastores configuration ────
+# ── Disable embedded databases ────
-# Disable auto-deployed PostgreSQL (we use external ddb for other services)
 postgresql:
   enabled: false
 
-# Enable Elasticsearch for visibility store (deployed to worker node, 2Gi/4Gi memory)
-elasticsearch:
-  enabled: true
-  scheme: http
-  host: elasticsearch-master-headless
-  port: 9200
-  version: v7
-  logLevel: error
-  auth:
-    enabled: false
-  indices:
-    visibility: temporal_visibility_v1
-  replicas: 1
-
-# Cassandra enabled for template validation; server.config overrides with actual hosts
-# Schema job template requires cassandra config to exist at top level
 cassandra:
-  enabled: true
+  enabled: false
-  replicas: 3
-  cluster:
-    seedSize: 1
-  port: 9042
-  # Was landing 2 of 3 replicas on talos-cp-1 -- spread across nodes so a
-  # single overloaded node can't stall gossip/join for the whole ring.
-  affinity:
-    podAntiAffinity:
-      preferredDuringSchedulingIgnoredDuringExecution:
-        - weight: 100
-          podAffinityTerm:
-            labelSelector:
-              matchLabels:
-                app: cassandra
-                release: temporal
-            topologyKey: kubernetes.io/hostname
 
-# ── Disable schema auto-setup (will initialize manually) ─────────
+elasticsearch:
+  enabled: false
+
+# ── Disable schema auto-setup ─────
 jobs:
   autoSetup:
     enabled: false
 
-# ── Temporal server config (Cassandra + Elasticsearch persistence) ──────────────────────────────
+# ── Temporal server config (PostgreSQL persistence) ──────────────────────────────
 server:
   replicaCount: 1
   jobService:
     enabled: false
-  # Spread frontend/history/matching/worker across nodes instead of letting
-  # them stack on whichever node the scheduler prefers (was: 59 of ~80
-  # cluster pods on talos-cp-1 alone). History's ringpop gossip join was
-  # timing out because 3 of its 4 peers sat on that overloaded node.
   affinity:
     podAntiAffinity:
       preferredDuringSchedulingIgnoredDuringExecution:
@@ -77,22 +40,33 @@ server:
       numHistoryShards: 512
       datastores:
         default:
-          # Cassandra for workflow history and events
+          # PostgreSQL for workflow history and events
-          driver: cassandra
+          driver: sql
-          cassandra:
+          sql:
-            hosts:
+            driver: postgres12
-              - "temporal-cassandra-0.temporal-cassandra.temporal"
+            host: ddb-cluster-rw.ddb.svc.cluster.local
-              - "temporal-cassandra-1.temporal-cassandra.temporal"
+            port: 5432
-              - "temporal-cassandra-2.temporal-cassandra.temporal"
+            database: temporal
-            port: 9042
+            user: temporal
-            keyspace: temporal
+            password: ""
-            user: user
+            maxConns: 20
-            password: ""  # Cassandra auth disabled in deployment
+            maxIdleConns: 10
-            replicationFactor: 3
+            maxConnLifetime: "1h"
-            consistency:
+            connectAttributes:
-              default:
+              tx_isolation: "READ-COMMITTED"
-                consistency: local_quorum
+        visibility:
-                serialConsistency: local_serial
+          # PostgreSQL for visibility store (workflow queries)
+          driver: sql
+          sql:
+            driver: postgres12
+            host: ddb-cluster-rw.ddb.svc.cluster.local
+            port: 5432
+            database: temporal_visibility
+            user: temporal
+            password: ""
+            maxConns: 20
+            maxIdleConns: 10
+            maxConnLifetime: "1h"
   service:
     type: ClusterIP
 
@@ -103,9 +77,6 @@ web:
     type: ClusterIP
 
 # ── Ingress ────────────────────────────────────────────────────────
-# Note: ingress is disabled here. Instead, we route via oauth2-proxy.
-# The ingress is applied separately as k8s/temporal/temporal-ingress-oauth2.yaml
-# which terminates TLS and routes to oauth2-proxy service.
 ingress:
   enabled: false