if ceph-osd takes longer than 1 seconds to unmount the objectstore, it
won't get a chance to fully persist all of its stuff to disk before
init-ceph kills it again.
in my case, `MemStore::_save()` fails to write $path/collections before
it gets killed, if it has more than 512 PGs.
Signed-off-by: Kefu Chai <kchai@redhat.com>
cmd=\"kill $signal \$pid\"
printf \"\$cmd...\"
\$cmd
- sleep 1
+ sleep 2
continue
done
fi"