Yes, we should not have past intervals in this case. However, we don't
need to crash the OSD; the ERR in clog is sufficient to fail the teuthology
run, and users in production don't actually care.
See: http://tracker.ceph.com/issues/20167
Signed-off-by: Sage Weil <sage@redhat.com>
derr << info.pgid << " required past_interval bounds are"
<< " empty [" << rpib << ") but past_intervals is not: "
<< past_intervals << dendl;
- assert(past_intervals.empty());
}
} else {
if (past_intervals.empty()) {