Created attachment 692720[details]
logs
Description of problem:
I had some tasks that became zombies and I moved my host to a different DC.
the tasks, although zombie, could not be cleaned until I reconnected the pool about a week later.
Version-Release number of selected component (if applicable):
si26.1
how reproducible:
Steps to Reproduce:
1. run several live storage migration tasks and restart vdsm
2. put host in maintenance
3. create a new pool with the same host
4. wait for the tasks to become zombie tasks
Actual results:
engine fails to clean the zombie tasks if we cannot find master domain
Expected results:
we should be able to clean the tasks from the db since the tasks are already zombie tasks.
Additional info: logs
The problem is that when we reach a zombi state, we should not depend on whether there is an active SPM for clearing this task because:
- it should be cleaned ... too old
- in case no spm exist .. those tasks will not appear in it's task list there
for no cancel task API will be used.
Created attachment 692720 [details] logs Description of problem: I had some tasks that became zombies and I moved my host to a different DC. the tasks, although zombie, could not be cleaned until I reconnected the pool about a week later. Version-Release number of selected component (if applicable): si26.1 how reproducible: Steps to Reproduce: 1. run several live storage migration tasks and restart vdsm 2. put host in maintenance 3. create a new pool with the same host 4. wait for the tasks to become zombie tasks Actual results: engine fails to clean the zombie tasks if we cannot find master domain Expected results: we should be able to clean the tasks from the db since the tasks are already zombie tasks. Additional info: logs