using watershed in apport-noui.service causes multiple instances of watershed to run

[Brian Murray]

I've modified /lib/systemd/system/apport-noui.service to contain the following:

ExecStart=/lib/udev/watershed /usr/share/apport/whoopsie-upload-all

With the hope that bug 1473562 will be resolved under systemd. (Note that using watershed works with upstart and /etc/init/apport-noui.conf.)

I've tested this by setting the PathExistsGlob to '/var/crash/*.nothing', then creating a bunch of crash files as in bug 1473562, switching the PathExistsGlob back to '/var/crash/*.crash' and running 'sudo systemcl daemon-reload'. After this I see the following in my journal.

journalctl --since "4 min ago" /lib/udev/watershed
-- Logs begin at Sat 2015-06-27 08:38:54 PDT, end at Tue 2015-07-21 15:10:21 PDT. --
Jul 21 15:10:07 impulse watershed[22068]: Collecting info for /var/crash/_bin_dash_10.32011.crash...
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash_10.32011.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_bin_dash_09.32011.crash already has info collected
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash_09.32011.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_bin_dash_04.32011.crash already has info collected
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash_04.32011.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_bin_dash_03.32011.crash already has info collected
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash_03.32011.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_bin_dash_01.32011.crash already has info collected
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash_01.32011.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_bin_dash.1000.crash already has info collected
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash.1000.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_bin_dash_08.32011.crash already has info collected
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash_08.32011.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_bin_dash_07.32011.crash already has info collected
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash_07.32011.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_bin_dash_05.32011.crash already has info collected
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash_05.32011.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_bin_dash_02.32011.crash already has info collected
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash_02.32011.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_bin_dash_06.32011.crash already has info collected
Jul 21 15:10:07 impulse watershed[22068]: Marking /var/crash/_bin_dash_06.32011.crash for whoopsie upload
Jul 21 15:10:07 impulse watershed[22068]: /var/crash/_usr_share_apport_whoopsie-upload-all.1000.crash already marked for upload, skipping
Jul 21 15:10:07 impulse watershed[22068]: All reports processed
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash_10.32011.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash_09.32011.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash_04.32011.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash_03.32011.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash_01.32011.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash.1000.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash_08.32011.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash_07.32011.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash_05.32011.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash_02.32011.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_bin_dash_06.32011.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: /var/crash/_usr_share_apport_whoopsie-upload-all.1000.crash already marked for upload, skipping
Jul 21 15:10:08 impulse watershed[22308]: All reports processed
....

So watershed is called a number of times equal to the number of .crash files that exist in /var/crash/.

[Brian Murray (brian-murray)]

I'm looking for a way to limit the number of instances of whoopsie-upload-all run on boot up in systemd and I'm not certain if this behavior is a bug or if I haven't found the right way to set this limit.

[Martin Pitt (pitti)]

Why do you need watershed in the first place? With the current package, the unit gets activated once if the PathExistsGlob triggers, and not re-triggered while it's running. Once it's done uploading (and thus systemctl status apport-noui.service gets back to "inactive (dead)", it will be re-triggered on new events. Isn't that what we want?

What's missing here?

[Launchpad Janitor (janitor)]

[Expired for systemd (Ubuntu) because there has been no activity for 60 days.]