The mailing list for listmasters using Sympa

[Youcef N Baouchi <address@concealed>]

Hi All,

We had an incident where a user sent an email to a list, but 20 out of 205 users did not receive the email. Sympa log shows that sympa accepted the email for 205 users:

/var/log/sympa.log-20161002.gz:Sep 29 09:07:04 listssympa-prod sympa_msg[27745]: info Sympa::Spindle::ToList::_twist() Message Sympa::Message <address@concealed> for Sympa::List <address@concealed> from address@concealed accepted (0.73 seconds, 9 sessions, 205 subscribers), message address@concealed, size=513293

When I checked the sendmail log I found that 185 people were emailed instead of 205 subscribed users. I was able to see the grouped users as well as the users that were tagged for VERP.  There seems to be a discrepency with what sympa "thinks" it gave to sendmail and what it actually sent. What's more concerning is that there is no error or bounce showing an issue. We only found this problem because people reported it and then we confirmed it via our IronPort applicance that showed it  did not find any record of sympa sending email to the 20 users and we further confirmed it via o365's message trace. The sendmail logs also show that 185 people were emailed instead fo 205.

This same list was emailed successfully before and after this event with all 205 users receiving the message. We were able to confirm this via logs.

 I also confirmed that these users were subscribed to the list at this time. All users that were subscribed are configure to receive mail. No user is set to "no mail."

This appears to be another bug that we found with the newest release. The other massive bug that we've been dealing with has been random crashes of bulk.pl with an "out of memory" error. I've been working with Soji on this bug, but we're still gathering more data. We're afraid that sympa does not scale as well as we thought.

Thanks,
Youcef