> Define low. 2 disks maxed out doing random I/O could show as little as 2 or
> 3 MB/sec in vmstat.
While the copy commands are not at 100% CPU it is about that low. It does go up at times, but not by much (maybe 5-7 MB/sec). Does a COPY command count as random I/O if the data it's reading is actually a memory FIFO and it's only reading data in for the current (not completed) transaction? The disks are completely dedicated to the database, so postgres is the only app using it.
> You say one CPU core is maxed out - what state is it mostly in at 100% -
> user, system, or wait?
I'll double-check, but I'm certain that it was user. The behavior didn't seem to change when I changed the table layout either.
- Phillip
On Wed, Jul 9, 2008 at 9:45 AM, Alan Hodgson <
ahodgson@simkin.ca> wrote:
On Wednesday 09 July 2008, "Phillip Sitbon" <
phillip@sitbon.net> wrote:
> I only have two fast SATA drives on software RAID, but that really isn't
> the issue- while the copy commands are going, disk activity is relatively
> low.
Define low. 2 disks maxed out doing random I/O could show as little as 2 or
3 MB/sec in vmstat.
You say one CPU core is maxed out - what state is it mostly in at 100% -
user, system, or wait?
--