Created attachment 1304409 [details] journalctl output Description of problem: Starting preliminary perf-scale testing on CRI-O. Have a 4 node cluster with 1 master, and 3 nodes all using CRI-O on RHEL (not Fedora). The test I am running is the nodeVertical test which creates a large number of pods. (https://github.com/openshift/svt/blob/master/openshift_scalability/nodeVertical.sh) After ~118 pods are created, no further pods are able to be scheduled. The relevant error seems to be: Jul 25 14:11:36 atomic-openshift-node[56515]: E0725 14:11:36.306490 56515 kubelet.go:1666] Failed to check if disk space is available for the runtime: failed to get fs info for "runtime": ImagesFsInfo: unknown runtime: remote Version-Release number of selected component (if applicable): cri-o-1.0.0.alpha.0-1.git91977d3.el7 containernetworking-plugins-0.5.2-2.el7 How reproducible: Always Steps to Reproduce: 1. Have OpenShift cluster using CRI-O 2. Create large number of pods (node vertical test) 3. After ~118 pods they will no longer be schedulable. Actual results: Pods cannot be scheduled Expected results: Pods continue to be created and scheduled successfully Additional info: CRI-O kubelet arguments match those found here: https://raw.githubusercontent.com/ashcrow/openshift-ansible/571de035357e6bd219ce62e66dc786fbdedc9b9c/roles/openshift_node/templates/node.yaml.v1.j2 journal log attached
That log isn't the cause here, I suspect something else is going on here. I'll have a look at logs soon.
This was a OCP issue, now working as expected with 3.6.171. There are plenty of errors in the logs including the ones posted in this issue. However they do not seem to be blockers as Antonio stated. Closing.