We were doing this on FULL, should also have done it
on PAUSERD and PAUSEWR, to avoid paused requests
staying that way forever.
Signed-off-by: John Spray <john.spray@redhat.com>
{
assert(rwlock.is_locked());
int flag = 0;
- if (osdmap_full_flag()) {
+ if (osdmap_full_flag()
+ || osdmap->test_flag(CEPH_OSDMAP_PAUSERD)
+ || osdmap->test_flag(CEPH_OSDMAP_PAUSEWR)) {
ldout(cct, 10) << "_maybe_request_map subscribing (continuous) to next osd map (FULL flag is set)" << dendl;
} else {
ldout(cct, 10) << "_maybe_request_map subscribing (onetime) to next osd map" << dendl;