From b33fdde5b74498b58b1cd036f7b759ba91438e78 Mon Sep 17 00:00:00 2001
From: Story Crater Bot <19826264+Riotpiaole@users.noreply.github.com>
Date: Mon, 13 Jul 2026 13:20:50 -0700
Subject: [PATCH] fix: temporal service config - add explicit ClusterIP
 services for history/matching

---
 k8s/temporal/temporal-values.yaml | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/k8s/temporal/temporal-values.yaml b/k8s/temporal/temporal-values.yaml
index 3d22de7..6c93688 100644
--- a/k8s/temporal/temporal-values.yaml
+++ b/k8s/temporal/temporal-values.yaml
@@ -32,6 +32,18 @@ cassandra:
   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) ─────────
 jobs:
@@ -43,6 +55,19 @@ 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:
+        - weight: 100
+          podAffinityTerm:
+            labelSelector:
+              matchLabels:
+                app.kubernetes.io/instance: temporal
+            topologyKey: kubernetes.io/hostname
   config:
     logLevel: "info"
     persistence: