Comutation error - cause was out of memory

Message boards : Questions and problems : Comutation error - cause was out of memory
Message board moderation

To post messages, you must log in.

AuthorMessage
Gordon Haverland

Send message
Joined: 25 Oct 16
Posts: 3
Canada
Message 82117 - Posted: 19 Oct 2017, 23:08:09 UTC

Hello.

I have been running a bunch of SETI and Einstein@Home jobs, and what had become more or less steady state was 5 SETI CPU jobs and 1 Einstein job split between CPU and GPU on a 8 core CPU.

And then I started an interactive perldb session in emacs which generated a ton of output. And the machine ran out of memory, and the kernel started killing jobs. Seamonkey got clobbered, as did a pile of SETI and Einstein jobs.

Is there a preferred way to handle this? Should I have stopped boinc-client before I started this? What if you don't know that something similar is going to chew up all of RAM and SWAP (I have 8 GB of RAM and 16GB of swap).

I think I was running 7.8.2 for boinc. Computer is running Devuan/Ascii (nominally the same as Debian/testing). I think the 4.12 kernel. Mesa is supplying compute abilities for GPU (RX-460), not that proprietary stuff from AMD.

Thanks.
ID: 82117 · Report as offensive
Profile Yavanius
Avatar

Send message
Joined: 19 May 15
Posts: 123
Antarctica
Message 82118 - Posted: 20 Oct 2017, 0:07:46 UTC - in response to Message 82117.  

7.8.3 is released. Gianfranco should have it on the PPA now. You can see if that makes any difference.

However, it's good to shut down things down in advance if you think you might be consuming a lot of RAM and are not sure how much will get gobbled up. If it's your first time, you might do it during a period in which you can monitor for a whiles.

Beyond that, you can delve into scripting something to monitor resources and shut down things that are running and possibly sound an alarm (which of course doesn't help if you're around) if things get critical.
ID: 82118 · Report as offensive
Gordon Haverland

Send message
Joined: 25 Oct 16
Posts: 3
Canada
Message 82130 - Posted: 20 Oct 2017, 15:33:29 UTC - in response to Message 82121.  

I looked over preferences again, and I don't see how the BOINC manager could have stopped me from ruining a bunch of tasks. I think the manager(or client) had no choice but to label the first job killed by the kernel in an attempt to manage RAM as computation error, but it would have been nice if the manager(or client) would not have kept trying to start new jobs (and subsequently have them fail) while the load was so high.

One of my machines is running 7.8.3, so I could update the 7.8.2 to that. Would that have saved me? :-)

I guess I will eventually have to come up with a list of programs which I should enumerate for BOINC, which can have big demands. I don't see that this is practical with emacs, as it can do so many things that are not resource hungry.

What might be nice is if a person could send SIGUSR to the boinc process to tell it something unusual was happening. Perhaps use SIGUSR like XON/XOFF (Control S/Control Q)?
ID: 82130 · Report as offensive

Message boards : Questions and problems : Comutation error - cause was out of memory

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.