From d31c67c68c6f13c817ad217183a7e24c6d2f7116 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Jord=C3=A0=20Polo?= <jorda.polo@amd.com>
Date: Wed, 26 Jun 2024 01:44:58 -0500
Subject: [PATCH] Add healthcheck to ensure docker service is working

---
 test/docker/slurm/compose.yaml | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/test/docker/slurm/compose.yaml b/test/docker/slurm/compose.yaml
index 4aff4211..7541f74d 100644
--- a/test/docker/slurm/compose.yaml
+++ b/test/docker/slurm/compose.yaml
@@ -36,6 +36,12 @@ services:
       - controller
     links:
       - controller
+    healthcheck:
+      test: curl --fail localhost:8000/metrics || exit 1
+      interval: 15s
+      timeout: 10s
+      retries: 5
+      start_period: 15s
 
 volumes:
   jobs_dir: