Test juju and network after reboot
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
juju-ci-tools |
Fix Released
|
High
|
Curtis Hovey |
Bug Description
Juju has a long history of bugs caused by rebooting a machine and/or its containers. CI could testing rebooting to catch regressions.
Rebooting any env and verify that juju reports all agents are ready is the simplest level of check. This verifies that after reboot jujud did restart and the the hosts network matches Juju's expectations and needs to call home.
Rebooting containers that are visible to each other and outside networks is more difficult to test, This case is only supported by juju 1.25 on aws with a feature-flag (lxc). Juju 2.0 supports this on maas 1.9 and 2.0 (lxd). Even though juju will report everything is healthy, the internetworking needs to be tested separately. It can be broken even when Juju reports all is well.
The container networking test only supports aws. When the feature-flag for aws was removed form juju 2.0, we lost the reboot aspect of testing. At the time the feature was removed, the jobs was failing because container networking was broken after the reboot, even though juju reported all was well.
Changed in juju-ci-tools: | |
status: | In Progress → Fix Released |
Leo and I developed a fix for this for trusty maas-2.0. We can update the jobs to use the new test.