From ef7f4bc98bc3012798b9bb5b51a5da98b2389e59 Mon Sep 17 00:00:00 2001 From: Wesley Ellis Date: Mon, 6 Jul 2026 15:00:00 -0400 Subject: [PATCH] Fix job pod memory limit multiplier typo (1204 -> 1024) The memory limit was computed as MB*1024*1204 instead of MB*1024*1024, inflating every job pod's effective memory limit by ~17.6% over the configured value (e.g. --job-pod-limits-memory=6656 produced ~7.8GiB instead of 6.5GiB). Note this effectively lowers job pod memory limits for existing deployments; jobs that relied on the accidental headroom may start OOMKilling after this lands. Co-Authored-By: Claude Fable 5 --- src/pkg/k8s.go | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/pkg/k8s.go b/src/pkg/k8s.go index 3efe7e15..f9caa209 100644 --- a/src/pkg/k8s.go +++ b/src/pkg/k8s.go @@ -218,7 +218,7 @@ func (s *JobRunner) getPodObject(identifier string, labels map[string]string, jo }, Limits: corev1.ResourceList{ corev1.ResourceCPU: *resource.NewMilliQuantity(s.jobPodConfig.CpuLimit, resource.DecimalSI), - corev1.ResourceMemory: *resource.NewQuantity(s.jobPodConfig.MemLimit*1024*1204, resource.BinarySI), + corev1.ResourceMemory: *resource.NewQuantity(s.jobPodConfig.MemLimit*1024*1024, resource.BinarySI), }, }, Env: s.getPodEnv(job.Variables),