ceph-disk systemd unit timesout too quickly
When there are more than a few disks, it is likely to take longer than five minutes.
http://tracker.ceph.com/issues/18060 proposed to reduce the lock contention but it does not really help because there is another global lock within ceph-disk right after that. It effectively is a noop.
http://tracker.ceph.com/issues/18740 is a better workaround because it allows to configure the timeout without modifying the systemd unit.
It would be simpler and have no undesirable side effect to just set the timeout to a very large value. There needs to be a timeout to prevent ceph-disk from hanging forever (that was the first incentive to set a timeout). But there is no good reason to set it to a value that is less than a few hours.