Infrastructure » sepia

With git commit https://github.com/ceph/ceph-qa-suite/commit/1749ba912aa8d4fd05edde45d311b9411b760609, the ceph-deploy suite started testing separate block devices for journals vs OSDs. Since each machine needs 3 OSDs (https://github.com/ceph/ceph-qa-suite/blob/master/suites/ceph-deploy/fs/tasks/cfuse_workunit_suites_blogbench.yaml#L13-L15), this means that a node must have 6 disks to run the suite. This makes the suite to big for plana nodes, which is making this suite very difficult to get through testing for Hammer release. The size of the suite has also more than quadrupled since firefly.

We could potentially cut the OSD count down to 2 per node (will this fit on plana then?), or maybe only do these override on a subset of the tasks.

Long term, there has been talk of re-writing the ceph-deploy suite. It doesn't really appear to be exercising the capabilities of ceph-deploy itself.