Re: Proposal: Log inability to lock pages during vacuum

On 12/18/14, 7:56 AM, Robert Haas wrote:
> On Wed, Dec 17, 2014 at 11:20 AM, Heikki Linnakangas
> <hlinnakangas(at)vmware(dot)com> wrote:
>> LOG: automatic vacuum of table "postgres.public.foo": index scans: 0
>> pages: 0 removed, 7256 remain, 0 pinned
>> tuples: 79415 removed, 513156 remain, 0 are dead but not yet
>> removable
>> buffer usage: 14532 hits, 6 misses, 6241 dirtied
>> avg read rate: 0.003 MB/s, avg write rate: 3.413 MB/s
>> system usage: CPU 0.00s/0.30u sec elapsed 14.28 sec
>>
>> I.e. this just says how many pages were pinned, without saying what was done
>> about them. That's not very meaningful to an average DBA, but that's true
>> for many of the numbers printed in vacuum verbose.
>
> That message is extremely confusing, to my eyes. If you want to say
> "pages: 0 removed, 7256 remain, 0 skipped due to pins", that would
> work for me, but just say "0 pinned" is totally wrong, because vacuum
> pinned every page in the table.

We have to decide on a tradeoff here. Either we end up with two different log messages (depending on scan_all) that require two different translations, or we end up with a generic message that isn't as clear.

The best option I can think of for the later is something like "failed initial lock attempt". That's the only thing that will be true in all cases.
--
Jim Nasby, Data Architect, Blue Treble Consulting
Data in Trouble? Get it in Treble! http://BlueTreble.com