Laggy slave on ZFS

We have 5 MySQL slaves running 5.0.x, two of them are solaris based, one of which has datadir on a ZFS volume for snapshot backups.

Of the 5 slaves, 4 keep up with sub-second replication lag. The fifth which has datadir on ZFS can not keep up, and falls behind the master nearly a second per second.

The ZFS filesystem’s recordsize is set to 16K to match InnoDB’s record size, and compression is turned on for the volume.

Anyone have any top-of-the-head suggestions as to what to look at?

I had a similar issue, I made sure the following two options were set false. Makes transaction recover a little more risky, but provides a big speed improvement.

innodb_flush_log_at_trx_commit=0
sync_binlog=0