Slow list

[Dan Mahoney]

On Tue, 5 Jul 2011, Chris Thompson wrote:

> On Jun 1 2011, Alan Clegg wrote:
> 
> > On 6/1/2011 7:16 AM, /dev/rob0 wrote:
> > > On Wed, Jun 01, 2011 at 09:54:04AM +0200, Jan-Piet Mens wrote:
> > > > > Does anyone else find the bind-users list to be very slow?
> [...]
> > I'll have operations take a look into what is causing the delay (it
> > doesn't happen on all mail to the list...)
> 
> Delays were very noticeable today for the bind-announce messages copied
> to bind-users. From the Received headers, the delay seems to happen
> between webster.isc.org and the next host: either mx.ams1.isc.org or
> mx.pao1.isc.org for different messages. In the worst case, there was
> delay of over 3 hours between webster receiving the message and passing
> it on.

We are working on this -- I believe we've discovered the problem 
internally, it seems to be a bad link between the handoff from mailman to 
postfix on a large list, exacerbated by several users whose domains return 
a SRVFAIL response.  (The irony here of course is that those users need 
the help the list could offer the most).

This seems to be because mailman asks postfix to do a "verify" (but not an 
SMTP VRFY) of the addresses as part of the VERP that it does.

One annoying thing that I should note is that removing those problem users 
and flushing the queues does NOT help.

-Dan