On Tue, Jun 14, 2016 at 03:06:11PM -0400, Jason J. Herne wrote: > On 06/14/2016 02:17 PM, Stefan Hajnoczi wrote: > > v4: > > * Add .pause()/.resume() callbacks to really quiesce during > > block_job_pause_point() [Paolo] > > * Add AioContext change support for backup block job > > * Tested drive_mirror + migration and drive_backup + reboot > > > > v3: > > * Push infrastructure down into blockjob.c so other jobs can reuse it > > [Stefan] > > * Tested with drive_mirror + migration [Stefan] > > > > v2: > > * Fam introduced the concept of a synchronous aio_poll() loop to quiesce > > the > > block job during detach > > > > When dataplane is enabled or disabled the drive switches to a new > > AioContext. > > The mirror and backup block jobs must also move to the new AioContext so > > that > > drive accesses are always made within its AioContext. > > > > This series extends the block job pause functionality so that detaching > > from an > > AioContext pauses the job and attaching to the new AioContext resumes the > > job. > > > > Pause points are added to the mirror job so that long I/O loops can yield > > for > > an AioContext switch. Other block jobs need pause points too but this can > > be > > done as a follow-up series. > > I just tested v4 on s390. It appears to fix our original problem without any > hiccups. > Thank you Stefan! :) Let me know if you need any more testing.
Thank you! Stefan
signature.asc
Description: PGP signature
