Comment 12 for bug 1459743

What I see on the environment provided is that an important detail was omitted in bug description:

If you try to run any OSTF test that requires a healthy ceph environment after step 2 (2 computes destruction, 2 nodes addition and env redeploy) you will receive:

"There are no compute nodes" failure.

So I see the bug as follows:

if you brutally destroy ceph cluster and try to atone for your sins by creating 2 nodes and redeploying env, the 2 new nodes won't be automatically picked up as computes. Any first request that requires glance's interaction with ceph (like PUT) will cause glance to hang due to the reasons described in details above (see cinder-like eventlet + librbd problem) and due to the absence of rados timeout by default.

My proposal is to mark the problem as invalid for 6.1 and schedule the glance_store bug fix for 7.0.