On Tue, Jun 4, 2013 at 5:57 AM, Federico Campoli <federico@brandwatch.com>wrote:
>
> I'm sorry, just guessing it could be a loop.
> The read remains stuck on the same segment.
> On my testbox I have at least 1 minute to 20 Mb/s.
> On the live system the peak is 124 Mb/s for 2 to 3 minutes without any
> progress in the wal reply.
>
> I've attached the part of postgresql's log with debug2 from my sandbox
> when that happens.
>
It looks like it is making progress, just very slowly. Basically I think
every WAL record that needs it to be replayed triggers the random read of
some data block which is not already cached on the standby.
> In warm standby everything is fine no lag at all.
>
OK, then I don't think I can reproduce it. The spiky replay I see is the
same whether the standby is hot or warm.
Cheers,
Jeff