From b3c86bb355a6cd46466a19b86a757f12280e2635 Mon Sep 17 00:00:00 2001
From: "Kevin H. Luu" <kevin@anyscale.com>
Date: Mon, 2 Dec 2024 18:09:55 -0800
Subject: [PATCH] Add retry for AMD jobs

---
 scripts/test-template-aws.j2 | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/scripts/test-template-aws.j2 b/scripts/test-template-aws.j2
index 718e31f..6068fe9 100644
--- a/scripts/test-template-aws.j2
+++ b/scripts/test-template-aws.j2
@@ -185,6 +185,8 @@ steps:
               limit: 2
             - exit_status: -10  # Agent was lost
               limit: 2
+            - exit_status: 1  # Machine occasionally fail
+              limit: 2
         agents:
           queue: amd-cpu
 
@@ -204,6 +206,15 @@ steps:
         {% else %}
         soft_fail: false
         {% endif %}
+        retry:
+          automatic:
+            - exit_status: -1  # Agent was lost
+              limit: 2
+            - exit_status: -10  # Agent was lost
+              limit: 2
+            - exit_status: 1  # Machine occasionally fail
+              limit: 2
+        
     {% endif %}
     {% endfor %}