Re: [BUGS] BUG #7642: Unchecked “stats_temp_directory” causes server hang to requests

From: Tianyin Xu <tixu(at)cs(dot)ucsd(dot)edu>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: pgsql-bugs(at)postgresql(dot)org
Subject: Re: [BUGS] BUG #7642: Unchecked “stats_temp_directory” causes server hang to requests
Date: 2012-11-07 19:17:08
Message-ID: CABBDWwdDvMw4OPhVH7MWG2WyqVQKYvxkPSR+f-W98D9KL9EPmQ@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-bugs

Thanks for the reply, Tom.

On Wed, Nov 7, 2012 at 10:53 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> tixu(at)cs(dot)ucsd(dot)edu writes:
> > I set the stats_temp_directory='' in my postgresql.conf file, with the
> > expectation that the directory would be under the data directory.
>
> Why would you think that's a good idea? It's already going to be under
> the data directory. The only reason to change this setting is if you're
> going to point to a ramdisk somewhere, which is going to need an
> absolute path.
>
>
I just changed my previous setting to the empty one. I thought it would be
under the data directory also. I got your point, I was stupid (I should
comment this one if I don't need it).

What I notice is that any wrong setting might cause latter problems. Since
many of us redirect the log messages into files, sometimes we do not notice
the problem immediately (at least for me). Maybe I'm too careless. For this
one, when I noticed the problem, the log file was filled up with the same
error message, and many requests stuck there...

> > I found pg has several unchecked user configurations which might cause
> > problems at run-time. I suggest to check it like what has been done for
> > “data_directory” to avoid latter problems. Here is the patches.
>
> This patch seems to have got mangled by your mailer, but in any case I
> don't really see the value. There are any number of ways to mess up
> settings like these, and only some of them are detectable by tests
> of this sort. Also, the proposed patch only checks at startup time;
> if we're going to install training wheels, that doesn't seem sufficient.
>
>
Yes, so I resend the previous mail.

I agree with you that users might make thousands of different mistakes, and
all the checkers are not complete. My personal opinion is to check as early
as possible, instead of finding errors at runtime. I really like the idea
of data directory checking and all the out-of-range checking of the numeric
values. What I envisioned is similar checking for semantic values. I don't
know whether it's possible (Sorry for my crappy patch).

btw, what is training wheels?

> regards, tom lane
>

--
Tianyin XU,
http://cseweb.ucsd.edu/~tixu/

In response to

Browse pgsql-bugs by date

  From Date Subject
Next Message Craig Ringer 2012-11-08 02:19:56 Re: duplicate key value violates unique constraint "tableName_pk"
Previous Message Tom Lane 2012-11-07 18:53:09 Re: BUG #7642: Unchecked “stats_temp_directory” causes server hang to requests