Page MenuHomePhabricator

MassMessage not delivering messages
Closed, ResolvedPublic

Description

I've tried to deliver some messages using MassMessage today. I first make some test messages to check if MassMessage was still working.

  • 11:08, 30 September 2019‎ ‎: 3 of 6 messages delivered (on meta, ar, fr), using this list
  • 12:48, 30 September 2019‎ : 0 of 8 messages delivered, using this list

We are really dependent on this tool, and not having it reliably delivering messages is really impacting Community-Relations work.

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptMon, Sep 30, 3:03 PM
Johan added a comment.Mon, Sep 30, 3:09 PM

Yeah I tried to send a message to to my test list to deliver Tech News but one message out of six made it so I don't think there's much point in trying right now.

Johan added a comment.EditedMon, Sep 30, 3:11 PM

Just to contextualise – among other things, this means that the weekly technical newsletter on which we rely to inform the communities about upcoming problems, changes, new features, feedback we need to get on tools being developed and so on can't be delivered. Basically, we're very limited in our ability to talk to the communities.

Joe added a subscriber: Joe.Mon, Sep 30, 3:22 PM

FWIW, according to the logs all jobs have executed correctly.

Johan added a comment.Mon, Sep 30, 3:41 PM

Does that mean that the jobs didn't include the undelivered messages?

Ok. I know what's happening, I'm sorry about that. This is a regression of https://gerrit.wikimedia.org/r/c/mediawiki/extensions/MassMessage/+/538742

The deduplication info provided by the default Job duplicate handler doesn't take the job domain into account, thus it deduplicates jobs with the same parameters across wikis.

Change 539899 had a related patch set uploaded (by Ppchelko; owner: Ppchelko):
[mediawiki/services/change-propagation@master] Take job domain into account when deduplicating.

https://gerrit.wikimedia.org/r/539899

Change 539899 abandoned by Ppchelko:
Take job domain into account when deduplicating.

Reason:
muscle memory

https://gerrit.wikimedia.org/r/539899

Mentioned in SAL (#wikimedia-operations) [2019-09-30T16:32:40Z] <ppchelko@deploy1001> Started deploy [cpjobqueue/deploy@79db711]: Take job domain into account for deduplication T234226

Mentioned in SAL (#wikimedia-operations) [2019-09-30T16:34:00Z] <ppchelko@deploy1001> Finished deploy [cpjobqueue/deploy@79db711]: Take job domain into account for deduplication T234226 (duration: 01m 17s)

This should be resolved now. Please test.

Johan added a comment.Mon, Sep 30, 4:43 PM

Successfully delivered to my test list of seven pages.

Pchelolo closed this task as Resolved.Mon, Sep 30, 4:43 PM
Pchelolo claimed this task.

Yay! Sorry about the regression.

Johan added a comment.Mon, Sep 30, 5:08 PM

Thanks for the quick fix. Much appreciated.

Thanks for the quick fix. Much appreciated.

Ditto! We really rely on this tool. So thank you for your great help!