log in

Advanced search

Message boards : Number crunching : Lots of RAM usage with this batch

1 · 2 · Next
Author Message
zombie67 [MM]
Avatar
Send message
Joined: 25 Jan 08
Posts: 89
Credit: 4,496,276
RAC: 0
Message 4430 - Posted: 17 Jun 2020, 13:11:17 UTC
Last modified: 17 Jun 2020, 13:11:17 UTC

I am seeing tasks using almost 6 gb of RAM each. Is that intentional? Most machines will choke on that.
____________
Reno, NV
Team: SETI.USA

Profile HK-Steve
Send message
Joined: 1 Nov 17
Posts: 3
Credit: 1,740,943
RAC: 0
Message 4431 - Posted: 17 Jun 2020, 13:20:53 UTC
Last modified: 17 Jun 2020, 13:20:53 UTC

Am seeing the same, ouch

Olivia L
Project administrator
Project developer
Send message
Joined: 26 May 20
Posts: 8
Credit: 0
RAC: 0
Message 4432 - Posted: 17 Jun 2020, 14:12:54 UTC
Last modified: 17 Jun 2020, 14:12:54 UTC

Apologies. There was an issue with how the job was setup, which resulted in many computational errors. The job was suspended while we fix the problems.

zombie67 [MM]
Avatar
Send message
Joined: 25 Jan 08
Posts: 89
Credit: 4,496,276
RAC: 0
Message 4433 - Posted: 17 Jun 2020, 14:21:03 UTC
Last modified: 17 Jun 2020, 14:21:03 UTC

Should we abort any tasks already in progress? Or let them complete?
____________
Reno, NV
Team: SETI.USA

Profile HK-Steve
Send message
Joined: 1 Nov 17
Posts: 3
Credit: 1,740,943
RAC: 0
Message 4434 - Posted: 17 Jun 2020, 17:55:30 UTC
Last modified: 17 Jun 2020, 17:55:30 UTC

I stopped new tasks, will finish the ones I have.
Hope we get some news on what is happening.

ChelseaOilman
Send message
Joined: 9 Jan 19
Posts: 5
Credit: 1,917,438
RAC: 0
Message 4435 - Posted: 17 Jun 2020, 18:59:11 UTC
Last modified: 17 Jun 2020, 18:59:11 UTC

I aborted all mine. It kept crashing my computers.

Profile Conan
Avatar
Send message
Joined: 24 Mar 09
Posts: 62
Credit: 1,137,257
RAC: 0
Message 4436 - Posted: 18 Jun 2020, 4:38:19 UTC
Last modified: 18 Jun 2020, 4:38:19 UTC

I can't abort just yet as I can't get back into my computer due to it not responding anymore. Lack of memory for such tasks. 6 cores and only 8 MB, with a few WUs using over 4GB each, has just about killed my computer.

Conan

matsu_pl
Send message
Joined: 11 May 20
Posts: 2
Credit: 55,290
RAC: 0
Message 4438 - Posted: 19 Jun 2020, 8:53:24 UTC
Last modified: 19 Jun 2020, 8:53:24 UTC

In a PC with 32 GB RAM those large WUs finish properly in approx. 5 hours. They were credited 700 - 1000 points.

Gunnar Hjern
Send message
Joined: 23 May 20
Posts: 1
Credit: 812,518
RAC: 0
Message 4448 - Posted: 23 Jun 2020, 23:16:17 UTC
Last modified: 23 Jun 2020, 23:16:17 UTC

I'm sorry but I'll have to abort my tasks and wait for the memory issue to get solved first.
Most my computers haven't got more than 1 Gb/core,
so I cannot crunch anything that takes more than 750 Mb/task!
(Except for some of the heaviest Rosetta tasks I haven't seen anything else yet
that have stopped my computers due to excessive memory requirements.)
I think 6Gb/task will be a showstopper for most crunchers, so I hope the issue can be solved soon!
//Gunnar

ProDigit
Send message
Joined: 15 Nov 19
Posts: 13
Credit: 722,245
RAC: 0
Message 4449 - Posted: 24 Jun 2020, 0:24:59 UTC
Last modified: 24 Jun 2020, 0:24:59 UTC

Seems to be a garbage collection issue.
I have 32GB of RAM, and 32 threads, and regularly see 10 threads or more waiting for memory.
When I restart the service, it loads all the WUs from before, plus the remaining empty cores, and still has about 50% of memory spare.
If you look into HTOP, it shows in the memory bar as a brown color (cache), and could easily be moved to swp.

ProDigit
Send message
Joined: 15 Nov 19
Posts: 13
Credit: 722,245
RAC: 0
Message 4450 - Posted: 24 Jun 2020, 17:18:11 UTC
Last modified: 24 Jun 2020, 17:21:42 UTC

For those who are using Linux, you'll be able to limit the amount of threads allocated to mindmodeling at a time, by being root, and creating or modifying the file
/var/lib/boinc/projects/mindmodeling.org/app_config.xml

In Windows, I presume the file to be in:
C:\ProgramData\BOINC\Projects\mindmodelling.org
But I'm not running windows, so I could be wrong.


<app_config>
<app>
<name>ccl_wrap_linOnly</name>
<max_concurrent>6</max_concurrent>
</app>
</app_config>


I had to use 'quotation' marks, as the 'code' marks don't seem to work properly on this forum.

In the above example, up to 6 threads are used.
I'd recommend to keep it below 50% of your CPU thread count.

Hal Bregg
Send message
Joined: 16 Sep 18
Posts: 22
Credit: 40,246
RAC: 0
Message 4451 - Posted: 25 Jun 2020, 6:10:35 UTC - in response to Message 4430.
Last modified: 25 Jun 2020, 6:10:35 UTC

I am seeing tasks using almost 6 gb of RAM each. Is that intentional? Most machines will choke on that.


I had a few tasks that were ok. Now I am stuck with one that is waiting for memory.
I guess I am out for now until I will be able to add some more RAM to my aging host.

Luigi R.
Send message
Joined: 11 Mar 14
Posts: 4
Credit: 169,489
RAC: 0
Message 4452 - Posted: 25 Jun 2020, 15:10:25 UTC
Last modified: 25 Jun 2020, 15:17:39 UTC

Ram usage of this task is huge.

Hal Bregg
Send message
Joined: 16 Sep 18
Posts: 22
Credit: 40,246
RAC: 0
Message 4453 - Posted: 25 Jun 2020, 18:54:29 UTC
Last modified: 25 Jun 2020, 18:57:55 UTC

I had run one WU run for slightly over 2 hours. Task seems to stuck at 9% so I suspended all work and re-enabled it straight away. WU finished within 7 seconds and reported as valid. CPU runtime was zero.

Sent 25 Jun 2020, 16:33:12 UTC Report deadline 30 Jun 2020, 16:33:12 UTC Received 25 Jun 2020, 18:44:03 UTC


I had another task that had run for hours and eventually was reported as valid with similar runtime.

Hal Bregg
Send message
Joined: 16 Sep 18
Posts: 22
Credit: 40,246
RAC: 0
Message 4454 - Posted: 26 Jun 2020, 18:33:26 UTC
Last modified: 26 Jun 2020, 18:33:26 UTC

Exactly same problem as mine was reported already

https://mindmodeling.org//forum_thread.php?id=1131


Could this be investigated, please?

zombie67 [MM]
Avatar
Send message
Joined: 25 Jan 08
Posts: 89
Credit: 4,496,276
RAC: 0
Message 4456 - Posted: 30 Jun 2020, 6:13:32 UTC - in response to Message 4432.
Last modified: 30 Jun 2020, 6:13:32 UTC

Apologies. There was an issue with how the job was setup, which resulted in many computational errors. The job was suspended while we fix the problems.


I see we have new tasks to crunch, and they are still requiring up to 6gb per task. Insanity. I thought this was going to be fixed?
____________
Reno, NV
Team: SETI.USA

Olivia L
Project administrator
Project developer
Send message
Joined: 26 May 20
Posts: 8
Credit: 0
RAC: 0
Message 4457 - Posted: 30 Jun 2020, 13:14:57 UTC - in response to Message 4456.
Last modified: 30 Jun 2020, 13:14:57 UTC

Apologies. There was an issue with how the job was setup, which resulted in many computational errors. The job was suspended while we fix the problems.


I see we have new tasks to crunch, and they are still requiring up to 6gb per task. Insanity. I thought this was going to be fixed?


One of our scientists is trying to submit a job with a couple of problems. We have decided to move his task to a resource with higher computing power. Unfortunately, this will probably mean that volunteers will not have a job to crunch on.

Mumps [MM]
Send message
Joined: 12 Apr 10
Posts: 3
Credit: 22,228,794
RAC: 0
Message 4458 - Posted: 30 Jun 2020, 21:06:01 UTC
Last modified: 30 Jun 2020, 21:06:01 UTC

Bummer. I was content to be helping out here. And I'd be happy enough to keep running that batch, seeing as so far I'd only logged two Error'ed units.

Oh well. Hopefully there's other work on its' way.

Hal Bregg
Send message
Joined: 16 Sep 18
Posts: 22
Credit: 40,246
RAC: 0
Message 4459 - Posted: 1 Jul 2020, 9:51:01 UTC - in response to Message 4457.
Last modified: 1 Jul 2020, 9:52:16 UTC

Apologies. There was an issue with how the job was setup, which resulted in many computational errors. The job was suspended while we fix the problems.


I see we have new tasks to crunch, and they are still requiring up to 6gb per task. Insanity. I thought this was going to be fixed?


One of our scientists is trying to submit a job with a couple of problems. We have decided to move his task to a resource with higher computing power. Unfortunately, this will probably mean that volunteers will not have a job to crunch on.


Would not be easier to change project settings to stop feeding WUs to users with not enough RAM?

morgan
Send message
Joined: 3 May 13
Posts: 6
Credit: 265,784
RAC: 0
Message 4460 - Posted: 2 Jul 2020, 17:41:57 UTC - in response to Message 4459.
Last modified: 2 Jul 2020, 17:41:57 UTC

Apologies. There was an issue with how the job was setup, which resulted in many computational errors. The job was suspended while we fix the problems.


I see we have new tasks to crunch, and they are still requiring up to 6gb per task. Insanity. I thought this was going to be fixed?


One of our scientists is trying to submit a job with a couple of problems. We have decided to move his task to a resource with higher computing power. Unfortunately, this will probably mean that volunteers will not have a job to crunch on.


Would not be easier to change project settings to stop feeding WUs to users with not enough RAM?


NO! I got enough Ram to run Mindmolding at 50%, (even 100%), i have never seen a wu over 4GB at use.
and combined with a other project; 50%-50%.then it is ok and safe for me to keep on here.

1 · 2 · Next

Message boards : Number crunching : Lots of RAM usage with this batch


Main page · Your account · Message boards


Copyright © 2022 MindModeling.org