Disk reads when using streaming replication?

[Dave Johansen <davejohansen@xxxxxxxxx>]

We have a Postgres 9.2 configuration with streaming replication. During periods when you're running nothing on the slave but the replication, we notice spikes in disk reads.

What could be the cause of this?

Are all of the WAL files written to disk and then read before being applied?

Or can something cause the WAL files to have to be written to disk and then read before being applied?

Thanks,

Dave