[Mailman-Users] Will slices ever become permanent?

[Jeff Taylor]

Well that's a shame that there hasn't been any more interest.  I have a 
file documenting the changes needed, which are pretty simple, so if any 
interest is expressed in this thread I'll share the details.  It could 
be that folks don't know the option exists, but you were a great help to 
me when the original problem came up, and this has worked great in an 
environment where servers can be rebooted without notice.


On 04/07/2015 01:22 PM, Mark Sapiro wrote:
> On 04/07/2015 08:09 AM, Jeff Taylor wrote:
>> I've been running the modified slice code (three slice servers plus one
>> backup server) for the past couple years, and generally speaking it has
>> run perfectly.  However every now and then I find that mailman is
>> suddenly not running on any of the servers, with no helpful information
>> showing in the log files.  And it always comes down to discovering that
>> an update for mailman was released which overwrote the slice code that I
>> had added to mailmanctl.
>
> I almost never apply local patches to a packaged software distribution.
> If I have a need to patch the code, I will generally install the
> software from source so that I have control over when and how it is
> upgraded and I don't have to worry that an automatic or semi-automatic
> software update will revert my changes.
>
>
>> This has been such a handy setup, I'm surprised that I haven't seen
>> comments from anyone else using the code.  Surely I'm not the only one
>> running multiple mailman servers?
>
> As far as I know, there is no actual documented procedure/patch for
> doing this outside the archives of this list.
>
>
>> Anyway, I was wondering if there are any plans to ever make this a
>> permanent part of mailman?  Since the QRUNNERS variable is only created
>> for this setup, it seems it would be an easy check to determine which
>> method a person is using, and run the appropriate code for each, so
>> there would be no impact on people running the single-server setup.  The
>> same is true for the backup server -- a single variable is defined ONLY
>> on that machine, so the necessary code would only run if that variable
>> were present.
>
> Actually, the QRUNNERS list is defined in Defaults.py and used in every
> installation, and slicing of queues to be shared across multiple runners
> is a standard feature. Your changes address enabling the multiple
> runners to run on different machines. That only slightly complicates
> making this a general feature, but no, there are no plans to do so.
>
> As you note above, there has been very little interest in a feature like
> this expressed on this list. In fact, the threads beginning at
> <https://mail.python.org/pipermail/mailman-users/2012-December/074537.html>
> and
> <https://mail.python.org/pipermail/mailman-users/2014-May/077068.html>,
> both initiated by you are the only things I recall on this subject in
> the last 7 years.
>
> There may be interest, but I haven't seen it so I haven't been motivated
> to do anything about it. Also, I'm not sure how many use cases there
> might be. I.e. you have N production servers plus a backup that only
> processes messages that haven't been picked up for a while by a
> production server. Will there be people who want similar load sharing
> with either zero or greater than one backup server. I.e., documenting
> the configuration options and actually configuring a particular
> installation is a more complex problem than patching the code.
>
> I suggest if you really want it, you submit it as a feature request at
> <https://bugs.launchpad.net/mailman/+filebug>, and tag it 'mailman3', as
> it's more likely to be implemented in MM 3 than in MM 2.1.
>