Description of problem: If restarting VDSM during live storage migration, there is a task in db (iveMigrateVmDisksParameters) that is not deleted: engine=# select task_id, action_type, status, task_type, vdsm_task_id, action_params_class from async_tasks; task_id | action_type | status | task_type | vdsm_task_id | action_params_class --------------------------------------+-------------+--------+-----------+--------------------------------------+------------------------------------------------------------------ b5658a22-dc22-40e1-8275-ab29289939e3 | 1011 | 2 | 3 | 7114d94d-8d33-442e-8fb8-c43a841d58d2 | org.ovirt.engine.core.common.action.LiveMigrateVmDisksParameters 9d6eee72-8b15-4b2d-bf45-e9e5691d728b | 211 | 2 | 5 | 00000000-0000-0000-0000-000000000000 | org.ovirt.engine.core.common.action.RemoveImageParameters (2 rows In vdsm there is no running tasks: [root@aqua-vds4 ~]# vdsClient -s 0 getAllTasks [root@aqua-vds5 ~]# vdsClient -s 0 getAllTasks This scenario cause that the Auto-generated snapshot is locked and the vms disk is locked as well and this is stayed that way until manual intervention. Version-Release number of selected component (if applicable): vdsm-4.14.7-1.el6ev.x86_64 rhevm-3.4.0-0.20.el6ev.noarch How reproducible: 100% Steps to Reproduce: 1. LSM vm disk 2. restart vdsm during LSM 3. Actual results: Explained above Expected results: should fail gracefully Additional info:
Hi Raz, * Can you please attach relevant vdsm/engine logs. * In which step of LSM the vdsm has been restarted?
Created attachment 910269 [details] vdsm and engine logs
Hi Daniel, I restarted VDSM right after the engine report: "User admin moving disk <disk_alias> to domain <domain_name>."
Daniel, this seems a lot like bug 1122639. Could that fix solve this one too?
(In reply to Allon Mureinik from comment #6) > Daniel, this seems a lot like bug 1122639. Could that fix solve this one too? Yes, seems like it. Couldn't reproduce on latest build (neither on dev nor QE environments). Moving to ON_QA.
verified
RHEV-M 3.5.0 has been released, closing this bug.