Re: [HACKERS] Continuous buildfarm failures on hamster with bin-check

On 04/18/2017 08:23 AM, Michael Paquier wrote:
> On Tue, Apr 18, 2017 at 9:13 PM, Andrew Dunstan
> <andrew.dunstan@2ndquadrant.com> wrote:
>> Yeah, but the way you have done it could also to lead to errors unless
>> you're very careful, as I found on axolotl (which died recently,
>> unfortunately). There I had to set the stats_temp directory to a
>> branch-specific name so a crash on branch A didn't leave stats picked up
>> by a run on branch B. I had a script that ran before every buildfarm run
>> that made sure the branch-specific directories existed on the tmpfs.
> I didn't consider that. Still hamster is only set to run master so
> that's not a problem. Running more branches would be too much for this
> little dear as it includes TAP tests.
>
>>>  but it is not possible to set up that with the
>>> TAP tests. I could always disable --enable-tap-tests on this machine
>>> but I don't think that this is a correct answer. Something that could
>>> be done is to use an environment variable to set up a base directory
>>> for all the nodes created in PostgresNode.pm, and use that for
>>> temporary statistics with a custom folder name. But that's not
>>> scalable if we want to enforce more parameter.
>> What more parameters do you want?
> Increasing wal_sender_timeout and wal_receiver_timeout can help in
> reducing the failures seen.


OK, but you're only talking about a handful of these, right?

Lets's say we have a bunch of possible environment settings with names
that all begin with "PG_TAP_" PostgresNode.pm could check for the
existence of these and take action accordingly, and you could set them
on a buildfarm animal in the config file, or for interactive use in your
.profile.

cheers

andrew

-- 
Andrew Dunstan                https://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services