Hide Forgot
Description of problem: Not deterministic behaviour for /distribution/virt/start. Once it fails once it passes. Version-Release number of selected component (if applicable): Version - 0.6.5 How reproducible: Time to time. Steps to Reproduce: See Jobs: PASS: https://beaker.engineering.redhat.com/recipes/115958 FAILS:https://beaker.engineering.redhat.com/recipes/116482 Actual results: Once it fails once it passes. Expected results: Always passes. Additional info:
Gurhan, Can you look into this for me?
I don't think this is a bug. What /distribution/virt/start does is to issue the start command to the guest. If it encounters error, it errors out. Domain starting and stopping are asynchronous events for libvirt, which means it issues the command to the guest and returns. In this instance (Recipe 116482), what's happening is that /distribution/virt/start is being called twice in the same job, on the same guest (tasks 1265671 & 1265674 ) and the second start job is being called immediately after a stop job. So the chances are that the guest is still running, though it might be in the process of being shutdown, so the virsh errors out since the guest is still running. Now, we can definitely either improve this, or have another test to ensure that the stop test waits for the guest to shutdown. In fact, it originally used to do that but logging into the serial console of the guest and waiting for the guest to shutdown but it caused out issues. Would you want such capability?
I supposed that probably the Guest is still shuting down so I tried rh-tests-distribution-virt-guest-reboot but it was failing too.
Dimitar Yordanov, Is this still an issue? Gurhan had made a nuber of changes in recent months. I am hoping that this one was eresolved. If not we will have to reassign it to someone willing to take over the virt/install task. Thanks, Jeff