lack of osd_per_node causes crash, even with use_existing set to true
If osd_per_node is unset, then cbt will crash mid way through the run, even with use_existing set to true.
I just came across this problem.
I've found that this was fixed in https://github.com/ceph/cbt/commit/1c831804ea3c24e5e09c060fbb4f0691369133d2. However, this problem came back and exists in the current master branch. Specifically, it is caused by the line: https://github.com/ceph/cbt/blob/master/cluster/ceph.py#L122
In fact, just grepping through the code, I'm not sure if that line is even needed since I don't see any obvious uses for "osd_count" field in that class. However, I can't be certain since I haven't try the "use_existing = False" features yet and I'm not familiar with this codebase.