Message boards :
Number crunching :
VM environment needed to be cleaned up.
Message board moderation
| Author | Message |
|---|---|
 Send message Joined: 29 Aug 05 Posts: 1001 Credit: 6,266,460 RAC: 113  |
Hi; running two CMS tasks in a Windows 10 box. Unfortunately, it's a "managed" machine and I have very limited admin rights to it. Over the weekend, my organisation installed patches and the machine rebooted. I doubt that BOINC was shut down cleanly. When I logged in on Monday, one VM restarted, got a CMS job, and closed the BOINC task when the job finished. The other got into some sort of state, with messages such as this:
I've tried several times to get the second task to run, including waiting for the running task to finish, closing BOINC and removing the BOINC VMs from the VBox manager before restarting. However, the second task always end up with the above message and shows as "Powered Off" in the VBox manager.  |
 Send message Joined: 15 Jun 08 Posts: 2401 Credit: 225,181,806 RAC: 123,896   |
1. Set BOINC to NNT and suspend all not yet started tasks 2. Let BOINC finish all running tasks, then shut down BOINC 3. Remove all .../slots/x/ (but don't remove /slots/) 4. Open VirtualBox Manager (use the BOINC user account) and remove "suspect" VM entries 5. From the VirtualBox manager menu start the Media Manager and remove "suspect" vdi entries 6. Restart BOINC, resume suspended tasks (staggered!) and allow fresh work |
|
Send message Joined: 29 Aug 05 Posts: 1001 Credit: 6,266,460 RAC: 113 |
Thanks! I'll try that tomorrow when my current task has finished. |
|
Send message Joined: 29 Aug 05 Posts: 1001 Credit: 6,266,460 RAC: 113 |
OK, that seems to have fixed it. Thanks again. |
|
Send message Joined: 2 May 07 Posts: 2090 Credit: 158,562,261 RAC: 129,138 
|
1. Set BOINC to NNT and suspend all not yet started tasks Point two: Suspending all CMS-Tasks (ONE after the OTHER! with a difference of a few seconds and waiting if suspended in Virtualbox). If they are all suspended. you can close Boinc, all went to saved. After reboot (for example Windows updates) you can start them all in the same way (ONE after the OTHER! with a difference of a few seconds) Eight CMS-Tasks without problems are running again! https://lhcathome.cern.ch/lhcathome/results.php?hostid=10593998 |
|
Send message Joined: 15 Jun 08 Posts: 2401 Credit: 225,181,806 RAC: 123,896 |
The question here was what to do to cleanup the vbox environment. Steps (1./2.) are the most reliable method to avoid dealing with a mix of clean and unclean slots in Step (3.). Step (3.) removes the slots without mercy, clean (should be empty now) as well as unclean content. If some tasks are just suspended they will immediately fail when BOINC restarts. |
©2024 CERN