On 5/10/13 1:32 PM, Simon Riggs wrote:
The timing window between the write and the sync is negligible and yet I/O would need to occur in that window and also be out of order from the order of the write, which is unlikely because an I/O elevator would either not touch the order of writes at all, or would want to maintain sequential order to avoid head movement, which is what we want. I guess we should add here "...with disks, maybe not with SSDs".
It's not really safe to make any assumptions about I/O elevators. Reordering gets done from the perspective of the last item written. When the previous write was at the logical end of the disk, it can just as easily re-order a queue of writes in the complete reverse order they were issued in.
The only way you can ever get a useful guarantee is when an fsync returns completion. Writes can effectively go out in a completely random order until that point. All you can rely on is throwing up a stop sign that says "tell me when all of them are done". In between those, you have no idea of the ordering.
-- Greg Smith 2ndQuadrant US g...@2ndquadrant.com Baltimore, MD PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.com -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
