On Mon, Dec 8, 2014 at 4:58 PM, Vincent de Phily <
vincent.deph...@mobile-devices.fr> wrote:
> On Monday 08 December 2014 10:17:37 Jeff Janes wrote:
> > On Mon, Dec 8, 2014 at 4:54 AM, Vincent de Phily <
>
> > I don't think that routine vacuums even attempts to update relfrozenxid,
> or
> > at lea
Vincent de Phily writes:
> What happens when vacuum is killed before it had time to go though the index
> with its dead-TID buffer ?
The next run will go through the index again, looking for those same TIDs
(and possibly more).
regards, tom lane
--
Sent via pgsql-gene
On Tuesday 09 December 2014 16:56:39 Tom Lane wrote:
> Vincent de Phily writes:
> > It reads about 8G of the table (often doing a similar number of writes,
> > but
> > not always), then starts reading the pkey index and the second index (only
> > 2 indexes on this table), reading both of them full
Vincent de Phily writes:
> It reads about 8G of the table (often doing a similar number of writes, but
> not always), then starts reading the pkey index and the second index (only 2
> indexes on this table), reading both of them fully (some writes as well, but
> not as many as for the table), w
On Tuesday 09 December 2014 01:58:11 Vincent de Phily wrote:
> On Monday 08 December 2014 10:17:37 Jeff Janes wrote:
> > You can `strace` for the lseek command to see which file handles it is
> > currently working on, and
> > use lsof to turn those into names. You want to look at where it is in th
On Monday 08 December 2014 10:17:37 Jeff Janes wrote:
> On Mon, Dec 8, 2014 at 4:54 AM, Vincent de Phily <
>
> > bloat, which I'd like to get back asap). Currently about 80% of the IO is
> > devoted to the vacuum process (on average throughout the day, as
> > extrapolated
> > from atop output).
>
On Mon, Dec 8, 2014 at 4:54 AM, Vincent de Phily <
vincent.deph...@mobile-devices.fr> wrote:
> Hi List,
>
> I have a "autovacuum: VACUUM ANALYZE public.some_big_table (to prevent
> wraparound)" that has been running for over 13 days. The process is
> consuming
> IO so I'm confident it isn't stuck,
