[Mailman-Developers] URGENT Please Help: Duplicates with 2.0 final

[Barry A. Warsaw]

>>>>> "TJ" == Todd Joseph <todd@connactivity.connactivity.com> writes:

    BW> The error your getting in the log file indicates that one of
    BW> the queued files is getting deleted out from under qrunner.
    BW> That could be because you're getting multiple qrunner
    BW> processes because of some locking failure, or because your
    BW> qrunner lock is getting stomped on.

    TJ> I kinda thought so.  Any ideas on what could cause that?

Are you running Mailman on an NFS mounted file system?  Locking
/shouldn't/ be susceptible to NFS issues, but it still isn't a
terribly good idea.

Another scenario is that you've cranked up your qrunner process
limits, but not appropriatedly adjusted your qrunner lock lifetime.
In that case, qrunner could be running for a long time, exceeding the
lock lifetime, and allowing a future cron started qrunner to break the
still running qrunner's lock.

The best way to see if this is happening is to edit cron/qrunner
around about line 274, where the LockFile instance is created.  Add an
argument "withlogging=1" to turn on lock file logging.  Then you'd
study the log to see if you get any "lock broken" messages.

    TJ> One more tidbit, I have set the "Action when critical or
    TJ> excessive bounces are detected" to "Do nothing" So, I do have
    TJ> a fair number of bogus addresses in my list.  My sendmail
    TJ> queue does get kinda large.  I've set it up this way out of
    TJ> paranoia in maintaining the integrity of the list (IE not
    TJ> removing valid email addresses).

I don't see how that could matter, even if it leaves bogus addresses
in your recipient lists.  Then again, I'll bet this isn't a well
tested option (the bouncer stuff is an area that I haven't had much
time to review).  Anything's possible, but it's unlikely if none of
the log files is indicating errors.

BTW, the default setting doesn't delete any bouncing emails from your
list -- it just disables them (currently exactly the same as setting
"no mail", but probably for 2.1 those will be distinguishable).

    TJ> OK, I'm bad.  I changed them in the Defaults.py file.  I tried
    TJ> to use the mm_cfg.py file, but ran into a problem (was a few
    TJ> months ago, don't remember what exactly happened).

I'd like to know about those, since changing mm_cfg.py is better.
Otherwise future upgrades could overwrite your changes.
    
    TJ> I've included a copy of my "Defaults.py" and "mm_cfg.py" at
    TJ> the bottom of this message for reference.

I'll take a look.

    BW> Which DELIVERY_MODULE are you using?

    TJ> SMTPDirect (would I be better off with Sendmail?)

Nope, you're using the right one.

    BW> Are you running a news gateway?

    TJ> No.

Okay, so it's not fork leakage (and there are fences for that one, but
you never know).  I'm leaning heavily toward qrunner contention.

-Barry