The last_decay in the msg might be 0 (utime_t()), so the eclipse time is very
large and the original value would be decayed to 0. We should set last_decay
to current time when decoding decay counter.
Fixes: http://tracker.ceph.com/issues/24440
Signed-off-by: Zhi Zhang <zhangz.david@outlook.com>
(cherry picked from commit
25166273764417258c9ad5bd9861d01875d25a1e)
Conflicts:
src/common/DecayCounter.cc
::decode(val, p);
::decode(delta, p);
::decode(vel, p);
+ last_decay = t;
DECODE_FINISH(p);
}