General

Profile

CONGMIN YIN

  • Registered on: 04/21/2020
  • Last connection: 08/31/2021

Issues

Projects

Activity

09/14/2021

08:25 AM rbd Bug #52277 (New): [pwl] IO hang when the single IO size * io_depth > cache size
08:17 AM rbd Bug #52599 (Fix Under Review): [pwl] flush requests are dispatched in advance
If the user sends _write1 write2 write3 write4 flush write5_, and the depth is 2, it may become _write1 write2 write3...

09/10/2021

02:00 AM rbd Bug #52566 (New): pwl ssd] assert in _aio_stop() during shutdown
After fix https://tracker.ceph.com/issues/52235 by moving finish_op to the end of call back func. If aio_thread call ...

09/06/2021

02:18 PM Ceph Revision a1a20041 (ceph): librbd/cache/pwl: don't clear next_sync_point_entry prematurely
In SyncPointLogOperation::clear_earlier_sync_point(),
sync_point->log_entry->next_sync_point_entry was prematurely se...
03:05 AM rbd Bug #52235: [pwl] deadlock on AbstractWriteLog::m_lock during shutdown

> Does the fix work? Were you able to reproduce the deadlock or the above segfault with https://github.com/ceph/ce...

09/02/2021

09:39 AM rbd Bug #52235: [pwl] deadlock on AbstractWriteLog::m_lock during shutdown
I find a segment fault(core dump). I think this is the same root cause as the deadlock issue. Dead lock is caused by ...

09/01/2021

08:16 AM rbd Bug #52235: [pwl] deadlock on AbstractWriteLog::m_lock during shutdown
I wrote a script to loop SSD Unitest, ran it dozens of times, and finally reproduced the deadlock problem. On my mach...
02:53 AM rbd Bug #52235: [pwl] deadlock on AbstractWriteLog::m_lock during shutdown
Ilya Dryomov wrote:
> FWIW while trying to reproduce this deadlock I also hit https://tracker.ceph.com/issues/52465....
02:15 AM rbd Bug #52258: [pwl] The write back time of cache is too long
> Linking https://github.com/ceph/ceph/pull/42775 that was supposed to address this for posterity.
In RWL mode, I ...

08/31/2021

07:01 AM rbd Bug #52258: [pwl] The write back time of cache is too long
> Is that always the case? I saw "write back 1G from cache" take a lot longer than "write 1G directly to the cluster...

Also available in: Atom