Re: danger of stats_temp_directory = /dev/shm

On 2013-08-19 12:47:05 -0700, Josh Berkus wrote:
> Tom,
> 
> > I note BTW that similar complaints could be lodged against the
> > log_directory setting.  We've not worried about that one too much.
> 
> Actually, it does happen that when you change log_directory on a reload,
> stuff takes an uneven amount of time to "cut over"; that is, there's a
> few seconds while you're writing to both logs at once.  Materially,
> though, this isn't a serious operational issue (the logs are known to be
> asynchronous), so beyond confusing newbies, it's not something we'd want
> to fix.

I think Tom was talking about the issue I noted on upthread which is
that two different clusters could write to the same
temp_stats_directory.
I've seen the logfile equivalent happen, but it was pretty easy to
diagnose...

Greetings,

Andres Freund

-- Andres Freund                       http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services