Innodb unexplained IOs

[Bob Bankay]

Running 4.0.13 on Sparc Sol 9. mysql_max.

We have a large table (10 -20GB) that has a lot of volatility. Occasionally in the past we have done reorgs by the usual methods and at one time increased the free space of the innodb files from 20GB to 50GB. So the table was very sparse.

Now we are see a lot of unexplained IOs (in iostat vs mytop) that do not correspond to any queries. Here is a mytop line that show physical IOs:

62.27 reads/s, 17532 avg bytes/read, 21.45 writes/s, 1.73 fsyncs/s

Here is a comparable iostat output for the raid10 volume:

Vol rd/s wr/s Rkb/s Wkb/s IOms Busy%
md40 162.3 19.8 10958.4 482.0 0.0 10.9 60.0 3 100

What is doing the extra 100 reads/sec? When mysql is freshly restarted the large number of reads/sec start without any queries being active. We are concerned because the volume is at 100% busy, we normally see it at 60% busy in iostat.

Any thoughts?

Bob