I've got a branch that will actually move that into the main wq. https://github.com/athanatos/ceph/tree/wip-recovery-wq More generally, PGs don't start recovering (and therefore are not in that queue) until they have reservations locally and on recovery targets -- those reservations *are* prioritized by amount degraded, etc. -Sam On Wed, Feb 3, 2016 at 12:44 AM, Gaurav Bafna <bafnag@xxxxxxxxx> wrote: > Hi Cephers, > > I am going through OSD Code flow. I noticed that the work queue for > normal read write operations is Prioritized Queue (op_shardedwq in > OSD.h), whereas the recovery work queue is FIFO with respect to PGs . > The PG which was first discovered to be "need to be recovered" will > be recovered first. There might be a PG which has more recovery work > to do , but was last in the queue. So in that case, it will lead to a > higher recovery time. > > I think it is FIFO assuming that the PG which needs more recovery , > must have lost many OSDs and should be well ahead in the queue. > > Should we have a recovery queue implemented as priority queue for the PGs ? > > Please let me know whether I am right or there is some problem in my > understanding. > > > -- > Gaurav Bafna > 9540631400 > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html > > -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
