** Description changed:
[impact]
services configured with both OnFailure= and Restart= directives will
both start the failure service(s), as well as restarting itself. The
documentation (and later versions) indicate that OnFailure= services
will only be started once the service actually does stop (due to
failure) once its restart limit has been exceeded.
[test case]
create service files as:
- ubuntu@lp1849261-b:/etc/systemd/system$ cat test.service
+ ubuntu@lp1849261-b:/etc/systemd/system$ cat test.service
[Unit]
OnFailure=failure.service
[Service]
ExecStart=/bin/sh -c '/bin/sleep 10; /bin/false'
Restart=always
- ubuntu@lp1849261-b:/etc/systemd/system$ cat failure.service
+ ubuntu@lp1849261-b:/etc/systemd/system$ cat failure.service
[Service]
ExecStart=/bin/true
-
perform daemon-reload, and then start test service:
- ubuntu@lp1849261-b:/etc/systemd/system$ sudo systemctl daemon-reload
+ ubuntu@lp1849261-b:/etc/systemd/system$ sudo systemctl daemon-reload
ubuntu@lp1849261-b:/etc/systemd/system$ sudo systemctl start test
wait for a while for the test service to exit - at least 10 seconds, as
that's what we configured in the service, although I have waited longer
in this example. Then check the journal for both test and failure
services:
ubuntu@lp1849261-b:/etc/systemd/system$ journalctl -b -u test
-- Logs begin at Fri 2019-12-13 00:52:50 UTC, end at Fri 2019-12-13 01:18:34
UTC. --
Dec 13 01:17:43 lp1849261-b systemd[1]: Started test.service.
Dec 13 01:17:53 lp1849261-b systemd[1]: test.service: Main process exited,
code=exited, status=1/FAILURE
Dec 13 01:17:53 lp1849261-b systemd[1]: test.service: Failed with result
'exit-code'.
Dec 13 01:17:53 lp1849261-b systemd[1]: test.service: Triggering OnFailure=
dependencies.
Dec 13 01:17:53 lp1849261-b systemd[1]: test.service: Service hold-off time
over, scheduling restart.
Dec 13 01:17:53 lp1849261-b systemd[1]: test.service: Scheduled restart job,
restart counter is at 1.
Dec 13 01:17:53 lp1849261-b systemd[1]: Stopped test.service.
Dec 13 01:17:53 lp1849261-b systemd[1]: Started test.service.
Dec 13 01:18:03 lp1849261-b systemd[1]: test.service: Main process exited,
code=exited, status=1/FAILURE
Dec 13 01:18:03 lp1849261-b systemd[1]: test.service: Failed with result
'exit-code'.
Dec 13 01:18:03 lp1849261-b systemd[1]: test.service: Triggering OnFailure=
dependencies.
Dec 13 01:18:04 lp1849261-b systemd[1]: test.service: Service hold-off time
over, scheduling restart.
Dec 13 01:18:04 lp1849261-b systemd[1]: test.service: Scheduled restart job,
restart counter is at 2.
Dec 13 01:18:04 lp1849261-b systemd[1]: Stopped test.service.
Dec 13 01:18:04 lp1849261-b systemd[1]: Started test.service.
Dec 13 01:18:14 lp1849261-b systemd[1]: test.service: Main process exited,
code=exited, status=1/FAILURE
Dec 13 01:18:14 lp1849261-b systemd[1]: test.service: Failed with result
'exit-code'.
Dec 13 01:18:14 lp1849261-b systemd[1]: test.service: Triggering OnFailure=
dependencies.
Dec 13 01:18:14 lp1849261-b systemd[1]: test.service: Service hold-off time
over, scheduling restart.
Dec 13 01:18:14 lp1849261-b systemd[1]: test.service: Scheduled restart job,
restart counter is at 3.
Dec 13 01:18:14 lp1849261-b systemd[1]: Stopped test.service.
Dec 13 01:18:14 lp1849261-b systemd[1]: Started test.service.
Dec 13 01:18:24 lp1849261-b systemd[1]: test.service: Main process exited,
code=exited, status=1/FAILURE
Dec 13 01:18:24 lp1849261-b systemd[1]: test.service: Failed with result
'exit-code'.
Dec 13 01:18:24 lp1849261-b systemd[1]: test.service: Triggering OnFailure=
dependencies.
Dec 13 01:18:24 lp1849261-b systemd[1]: test.service: Service hold-off time
over, scheduling restart.
Dec 13 01:18:24 lp1849261-b systemd[1]: test.service: Scheduled restart job,
restart counter is at 4.
Dec 13 01:18:24 lp1849261-b systemd[1]: Stopped test.service.
Dec 13 01:18:24 lp1849261-b systemd[1]: Started test.service.
Dec 13 01:18:34 lp1849261-b systemd[1]: test.service: Main process exited,
code=exited, status=1/FAILURE
Dec 13 01:18:34 lp1849261-b systemd[1]: test.service: Failed with result
'exit-code'.
Dec 13 01:18:34 lp1849261-b systemd[1]: test.service: Triggering OnFailure=
dependencies.
Dec 13 01:18:34 lp1849261-b systemd[1]: test.service: Service hold-off time
over, scheduling restart.
Dec 13 01:18:34 lp1849261-b systemd[1]: test.service: Scheduled restart job,
restart counter is at 5.
Dec 13 01:18:34 lp1849261-b systemd[1]: Stopped test.service.
Dec 13 01:18:34 lp1849261-b systemd[1]: Started test.service.
-
ubuntu@lp1849261-b:/etc/systemd/system$ journalctl -b -u failure
-- Logs begin at Fri 2019-12-13 00:52:50 UTC, end at Fri 2019-12-13 01:19:15
UTC. --
Dec 13 01:17:53 lp1849261-b systemd[1]: Started failure.service.
Dec 13 01:18:03 lp1849261-b systemd[1]: Started failure.service.
Dec 13 01:18:14 lp1849261-b systemd[1]: Started failure.service.
Dec 13 01:18:24 lp1849261-b systemd[1]: Started failure.service.
Dec 13 01:18:34 lp1849261-b systemd[1]: Started failure.service.
Dec 13 01:18:44 lp1849261-b systemd[1]: Started failure.service.
Dec 13 01:18:55 lp1849261-b systemd[1]: Started failure.service.
Dec 13 01:19:05 lp1849261-b systemd[1]: Started failure.service.
Dec 13 01:19:15 lp1849261-b systemd[1]: Started failure.service.
-
note that the 'failure' service should not have been started at all.
[regression potential]
this changes how service failures are handled, specifically when
OnFailure= services are started. However, this changes the behavior to
be in line with the documentation, and the current behavior is unlikely
to be anyone's desired behavior - both starting the failure service
*and* restarting the failed service. However, this will cause a
regression for anyone *expecting* both the OnFailure= service(s) to be
started, as well as the failing service to be restarted.
Any other regression would likely occur during the state change of a
unit, most likely during stop/exit of a service.
[scope]
- This is needed only for Bionic and Xenial.
+ This is needed only for Bionic and Xenial (but marking WontFix for
+ Xenial).
The patches to fix this:
7f66b026bbdda6bd3ddf7f617bf2c3a475e4a665
2ad2e41a72ec19159c0746a78e15ff880fe32a63
are included already in Disco and later.
+
+ For Xenial, I'm marking this as WontFix because it has additional
+ backporting requirements; at least commit deb4e7080db and possibly more.
+ Without anyone specifically requesting this for Xenial, I think the
+ possibility for regression in Xenial due to the larger backport is too
+ high to proactively apply it.
[other info]
original description:
---
Within Ubunutu 18.04 systemd 237-3ubuntu10.31: amd64 arm64 armhf i386
ppc64el s390x is available,
The pull request for 240 is now requested to backport to 237 for Ubuntu
18.04, which is integrated within the IBM product.
Pull request: https://github.com/systemd/systemd/pull/9158
** Tags removed: xenial
** Changed in: systemd (Ubuntu Xenial)
Status: In Progress => Won't Fix
** Changed in: systemd (Ubuntu Xenial)
Assignee: Dan Streetman (ddstreet) => (unassigned)
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1849261
Title:
Update systemd for ubuntu 18.04 with fix for interaction between
OnFailure= and Restart=
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu-z-systems/+bug/1849261/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
