Starting multiple instances on SOC with integrated Ceph cluster fails

This document (7022468) is provided subject to the disclaimer at the end of this document.

Environment

SUSE OpenStack Cloud 6

Situation

Environment where external Ceph cluster is integrated into SUSE OpenStack Cloud may not be able to start a batch of instances. Following error will appear in nova-compute.log on compute nodes:

2017-11-30 20:37:37.928 13683 ERROR nova.compute.manager [instance: 152e8ba0-4830-485d-a27c-8d126d0cf429] Command: rbd import --pool vms /var/lib/nova/instances/_base/afa3b11577fe146b7c581ea13dd70885a7b490f5 152e8ba0-4830-485d-a27c-8d126d0cf429_disk --image-format=2 --id cinder --conf /etc/ceph/ceph.conf

Resolution

Change the TasksMax value for openstack-nova-compute to "Infinity":

# systemctl set-property --runtime openstack-nova-compute.service TasksMax=infinity

Verify the value has been adjusted:

# cat /sys/fs/cgroup/pids/system.slice/openstack-nova-compute.service/pids.max
max

Cause

The default configuration sets the limit of processes to 512 for service openstack-nova-compute:

# cat /sys/fs/cgroup/pids/system.slice/openstack-nova-compute.service/pids.max
512

In certain configurations with many OSDs and large images this value may not be sufficient and 'rbd import' will fail producing a core file. Issue has been reported to engineering and maintenance update will correct the settings.

Additional Information

Disclaimer

This Support Knowledgebase provides a valuable tool for SUSE customers and parties interested in our products and solutions to acquire information, ideas and learn from one another. Materials are provided for informational, personal or non-commercial use within your organization and are presented "AS IS" WITHOUT WARRANTY OF ANY KIND.