On Wed, 7 Jul 2004, tonix (Antonio Nati) wrote:

> At 06/07/2004 06/07/2004 -0400, you wrote:
> >Hi,
> >
> >I believe there was some discussion about this some time ago, but recent
> >events have made me think of this again...
> >
> >The "standard" chkuser patch that vpopmail uses (see Bill Shupp's update
> >to the original: http://www.shupp.org/patches/chkuser-0.6.mysql.patch) has
> >some rather nasty behaviours.
> Probably you have not read carefully previous posts on this topic.

Yeah, I should have read the patch before posting. :)  Sorry Antonio.

> The standard chkusr patch uses standard vpopmail calls, that do NOT handle
> return status from DB operations (Bill's version just include some Makefile
> changes, in order to semplify compilation, code is untouched), and this
> happens for every DB.
> So the problem is not inside chkusr, but inside the vpopmail library.

Yeah, and it's pretty nasty...  In digging around on various mailing lists
I'm finding more and more people that either have a problem with 550's or
just discovered after reading my posts that they did.

There's some larger issue here that involves a moderately loaded machine
running both mysql and qmail (and incidentally, vpopmail) having trouble
getting some sql queries out in time.  The mysql people (mostly Zawodny)
don't think it's a bug/problem with mysql but that qmail can easily swamp
a system to the point that mysql bogs down, even with all the tables
cached in memory.

For most vpopmail operations, it's not a big deal; an occasional login
failure or mail being deferred.  But on the chkusr side, a mysql burp
leads to rejected mail.

> This is a known problem, that will be resolved as vpopmail will
> integrate such checks (I've been told these checks on DB connects are
> going to be put inside vpopmail CVS).

I hope that gets backported to 5.4.x, sounds like a good fix.

> I'm preparing chkuser 2.0, that will integrate all these changes, and
> will improve a lot of other things.

Excellent.  I also have a coworker looking at the patch to see if he can
build a workaround.  I'm also considering just changing the patch to
return a temporary failure.  Considering most of what chkusr blocks is
spam, why not let it queue on the remote end?  Nasty, but oddly



> Cheers,
> Tonino
> >Thanks,
> >
> >Charles
> ------------------------------------------------------------
>          [EMAIL PROTECTED]            Interazioni di Antonio Nati
>     http://www.interazioni.it      [EMAIL PROTECTED]
> ------------------------------------------------------------

Reply via email to