Yes I separate out as follows:
PGDATA + 1 TABLE which needs to be cached (also workaround CLOG read
problem)
LOGS
DATABASE TABLES
DATABASE INDEX
to get a good view of IOs out
I have full_page_writes=off in my settings
I dont see spikes of increase on WAL during checkpoints (maybe due to my
setting) but the constant load which is in the range of about
2-2.5MB/sec which is not low but my load is high.
In my current run I do have async wal on with wal_writer_delay=100ms and
commit_delay off.
-Jignesh
Heikki Linnakangas wrote:
> Jignesh K. Shah wrote:
>> Since its really writes that I am having trouble.. the auto vacuum
>> message tells me 11 pages were removed and so many tuples were
>> removed.. I am guessing its writes.
>
> Do you keep track of I/O to WAL and data separately? WAL bandwidth
> will spike up when a checkpoint starts, because of full page writes.
>