ObjBencher::aio_bench sets data.finished to num_objs but it should be
set to zero since it reflects the number of objects read /
written. That's what the read bench do and the write bench should do the
same. It does not cause a problem, it is just confusing.
Signed-off-by: Loic Dachary <ldachary@redhat.com>
data.object_size = object_size;
data.in_flight = 0;
data.started = 0;
- data.finished = num_objects;
+ data.finished = 0;
data.min_latency = 9999.0; // this better be higher than initial latency!
data.max_latency = 0;
data.avg_latency = 0;
pthread_create(&print_thread, NULL, ObjBencher::status_printer, (void *)this);
lock.Lock();
+ data.finished = 0;
data.start_time = ceph_clock_now(cct);
lock.Unlock();
for (int i = 0; i<concurrentios; ++i) {