Skip to content

Commit dafdbe9

Browse files
committed
fix(nccl-test): Add explicit DNS policy to all-reduce deployment
- Add dnsPolicy: ClusterFirst to init container spec - Add dnsPolicy: ClusterFirst to main container spec - Ensures consistent DNS resolution behavior across pod lifecycle stages
1 parent f3ca2c9 commit dafdbe9

1 file changed

Lines changed: 2 additions & 0 deletions

File tree

Container-Root/eks/deployment/nccl-test/all-reduce.yaml-template

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -47,6 +47,7 @@ spec:
4747
template:
4848
spec:
4949
restartPolicy: OnFailure
50+
dnsPolicy: ClusterFirst
5051
containers:
5152
- image: ${IMAGE_URI}
5253
imagePullPolicy: Always
@@ -99,6 +100,7 @@ spec:
99100
tolerations:
100101
- { key: nvidia.com/gpu, operator: Exists, effect: NoSchedule }
101102
- { key: vpc.amazonaws.com/efa, operator: Exists, effect: NoSchedule }
103+
dnsPolicy: ClusterFirst
102104
containers:
103105
- image: ${IMAGE_URI}
104106
imagePullPolicy: Always

0 commit comments

Comments
 (0)