You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When I launch a container using nomad-driver-containerd, and it exceeds its limits, cgroups are not applied and the container doesn't get OOM killed. To give a comparison between docker and nomad-driver-containerd driver:
When stress.nomad exceeds 500 Mhz of CPU or 256 MB of memory, it's OOM killed.
However when I launch the same job (stress.nomad) using nomad-driver-containerd it keeps running and doesn't get OOM killed.
In the case of docker driver, IIUC docker is managing the cgroups for the container.
The question probably is, how does nomad manage resource constraints (cgroups) on workloads launched by other drivers e.g. QEMU, Java, exec, etc.
Does nomad apply/manage cgroups at the orchestration level?
The text was updated successfully, but these errors were encountered:
When I launch a container using
nomad-driver-containerd
, and it exceeds its limits, cgroups are not applied and the container doesn't get OOM killed. To give a comparison betweendocker
andnomad-driver-containerd
driver:stress.nomad
When
stress.nomad
exceeds 500 Mhz of CPU or 256 MB of memory, it's OOM killed.However when I launch the same job (
stress.nomad
) usingnomad-driver-containerd
it keeps running and doesn't get OOM killed.In the case of docker driver, IIUC
docker
is managing thecgroups
for the container.The question probably is, how does nomad manage resource constraints (cgroups) on workloads launched by other drivers e.g. QEMU, Java, exec, etc.
Does nomad apply/manage cgroups at the orchestration level?
The text was updated successfully, but these errors were encountered: