From 44f0e0c00726c405753e51b50c9589f22c7b841a Mon Sep 17 00:00:00 2001
From: Yang Wang <elainewy@meta.com>
Date: Wed, 29 Apr 2026 18:23:20 -0700
Subject: [PATCH] Enable Alloy clustering and increase memory limit for prod
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

1. Add alloy.clustering.enabled=true — tells the helm chart to create
   a headless service and pass --cluster.join-addresses, enabling peer
   discovery between replicas. DPM drops from ~2x to ~1x per series.

2. Increase memory limit from 1Gi to 2Gi — prod has more scrape targets
   than staging, and clustering adds memberlist overhead. Without this,
   prod Alloy pods OOMKill.

No changes to controller type (stays as Deployment), deploy.sh,
or smoke tests.

Authored with Claude.
---
 osdc/modules/monitoring/helm/alloy-values.yaml | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/osdc/modules/monitoring/helm/alloy-values.yaml b/osdc/modules/monitoring/helm/alloy-values.yaml
index 67c1002b..ec6ec731 100644
--- a/osdc/modules/monitoring/helm/alloy-values.yaml
+++ b/osdc/modules/monitoring/helm/alloy-values.yaml
@@ -21,13 +21,15 @@ controller:
       effect: NoSchedule
 
 alloy:
+  clustering:
+    enabled: true
   resources:
     requests:
       cpu: 250m
       memory: 512Mi
     limits:
       cpu: "2"
-      memory: 1Gi
+      memory: 2Gi
   configMap:
     content: |-
       // Discover and scrape all ServiceMonitor targets across namespaces.