NOTIC: Keep more failed NCCL benchmark jobs in the history instead of…

… successful ones
nebius · Jan 12, 2025 · c9855b7 · c9855b7
1 parent 77a1ce1
commit c9855b7
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/helm/slurm-cluster/values.yaml b/helm/slurm-cluster/values.yaml
@@ -99,9 +99,9 @@ periodicChecks:
     # CronJob timeout in seconds. By default, equals to 30 min
     activeDeadlineSeconds: 1800
     # Number of successful finished jobs to retain
-    successfulJobsHistoryLimit: 24
+    successfulJobsHistoryLimit: 3
     # Number of failed finished jobs to retain
-    failedJobsHistoryLimit: 3
+    failedJobsHistoryLimit: 24
     # NCCL test settings
     ncclArguments:
       # Minimum memory size to start NCCL with