Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
World Community Grid Forums
Category: Support Forum: BOINC Agent Support Thread: Clean decommisioning |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 12
|
Author |
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I work in a performance lab environment where we are frequently going to be decommissioning a machine (with BOINC agent), putting a customer workload on it (clean OS build with no BOINC agent), at the end of the engagement we destroy (delete) the customer data, overwriting with a clean OS install + BOINC agent.
We can usually get a few hours to a day notice before needing to clean off the install with BOINC agent. Generally, more than enough to allow a work unit to run to completion. Is there some way we can set the agent to complete the current work unit, raise an SNMP alert and shut down? |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi,
Yes there are several ways to do this to include forcing to report the final task immediately. 1) boinccmd instructions (from terminal/DOS window) to stop fetching new work and close the client at last task, which could be batch scripted to copy this to any new temp install. See http://boinc.berkeley.edu/wiki/Boinccmd 2) several cc_config.xml <options> lines that stop the client after the last task and force immediate completed task reporting. See http://boinc.berkeley.edu/wiki/Cc_config.xml (would though require restarting client). Actually, the newest v7 clients are designed to report completed tasks immediately when it knows it's 30 minutes before scheduled network interruption or computing interruption... then it's as simple as modding the local preferences computing/network scheduling values for the day/time to close and to set the activity menu options to run/network based on preferences (default). Recommended is to also set a default setup with the "Connect about every..." / "Minimum work buffer" and the "(Maximum) Additional work buffer" to zero days. This way the client only has running tasks on it and does not do a prefetch of new work before the current jobs finish. Normally an out-of-the-box client maintains a little buffer of about 6-8 hours work. There are some config features we do not like to spell out on the forums... they are intended only for those that re-image systems on a frequent basis [college classrooms for instance], or in your case doing burn in of systems that later move to production or are for delivery to customers. Let us know if you need more help or email contact WCG support@worldcommunitygrid.org for "personalized" support. |
||
|
pramo
Veteran Cruncher USA Joined: Dec 14, 2005 Post Count: 703 Status: Offline Project Badges: |
Huh.
----------------------------------------Nice 18 day total:) 19:356:03:39:05 (#4,277) |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: |
I have long thought that there should be a simple way to stop after the current work unit is completed. The Folding@home icon has a "Finish" option (when you right-click on it) that allows the current work unit to compete, but then no more. It is frequently useful.
|
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
How about the "No new tasks" button, Jim1348? KISS enough for me and this one in version 7 client even takes to reporting all completed tasks immediately. Those in a rush could abort any non-started tasks too. Nobody minds in these temp client situations such as with Deepfreeze.
|
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: |
How about the "No new tasks" button, Jim1348? KISS enough for me and this one in version 7 client even takes to reporting all completed tasks immediately. Those in a rush could abort any non-started tasks too. I use that too, but if you are running CEP2 and need to reboot as soon as possible without losing work, then you really need to be able to stop at the end of the current work unit. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Think there's a boinc.exe parm for that too... unload when current task is done. http://boinc.berkeley.edu/dev/forum_thread.php?id=7807&postid=45415
In DOS window or terminal type then enter: boinc --exit_after_finish Exits BOINC right after finishing a job. Of course if more than 1 CEP2 job runs [science not advised for temp burn in clients... appropriate for long term dedicated, 24/7 devices], then you're bound to loose more when the others happen to be on job step 3 or 13. (Hey I really could be one of the few reading manuals... knowledge is power, but letting know that you know is not always advisable, else you get pestered for every little nit bit and never get to do your own job. Days I will just in short refer to the FAQ index are such moments ;o) |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: |
(Hey I really could be one of the few reading manuals... knowledge is power, but letting know that you know is not always advisable, else you get pestered for every little nit bit and never get to do your own job. Days I will just in short refer to the FAQ index are such moments ;o) If there were a simple check box for it, you would lose both the pestering and the power. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
(As if I had any at all in BOINC world... ;P)
|
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: |
(As if I had any at all in BOINC world... ;P) We think that you are all-knowing. That should be good for something. |
||
|
|