Re: pg_dump issues

Hi Andrew,

On 10/01/2011 09:46 PM, Andrew Dunstan wrote:
> 
> 
> On 10/01/2011 05:48 PM, Joe Abbate wrote:
>> On 10/01/2011 05:08 PM, Andrew Dunstan wrote:
>>> There is also this gem of behaviour, which is where I started:
>>>
>>>     p1                p2
>>>     begin;
>>>     drop view foo;
>>>                        pg_dump
>>>     commit;
>>>                        boom.
>>>
>>> with this error:
>>>
>>>     2011-10-01 16:38:20 EDT [27084] 30063 ERROR:  could not open
>>>     relation with OID 133640
>>>     2011-10-01 16:38:20 EDT [27084] 30064 STATEMENT:  SELECT
>>>     pg_catalog.pg_get_viewdef('133640'::pg_catalog.oid) AS viewdef
>>>
>>> Of course, this isn't caused by having a large catalog, but it's
>>> terrible nevertheless. I'm not sure what to do about it.
>> Couldn't you run pg_dump with --lock-wait-timeout?
>>
> 
> How would that help? This isn't a lock failure.

I misinterpreted the error.  I have confirmed the behavior you see.  I'm
somewhat surprised there appears to be no ability to lock a database
exclusively for something like pg_dump/pg_restore, so you're not
surprised by concurrent activity against the catalogs.  I'm guessing the
assumption is that MVCC will take care of that?

Joe