Before we switching the mds_lock to a fair mutex, the Finisher could
always successfully acquire the mds_lock if there has a number of
contextes finished. So it may take a long time to finish and will
delay kicking the heartbeat.
While even we have fix the mds_lock issue, it may still will happen
in case the complete() itself will take a long time to finish or
there has a number of contextes keep comming, which could make the
Finisher thread keep running.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
ceph_assert(mds != nullptr);
ceph_assert(ceph_mutex_is_locked_by_me(mds->mds_lock));
dout(10) << "MDSContext::complete: " << typeid(*this).name() << dendl;
+ mds->heartbeat_reset();
return Context::complete(r);
}