Boinc 100% disk activity, locks up

Message boards : BOINC Manager : Boinc 100% disk activity, locks up
Message board moderation

To post messages, you must log in.

AuthorMessage
tbzep

Send message
Joined: 29 Jan 17
Posts: 9
United States
Message 103374 - Posted: 2 Mar 2021, 21:00:33 UTC

I run Einstein on my Boinc manager. In the last couple weeks, it has started causing 100% disk activity and locks the computer up. As far as I know, nothing has been changed to cause this. I have purged work units, updated Windows, and finally, I did a reinstall/repair of Boinc. Nothing helped. When I start it, I get 3-5 minutes of good production with very little disk activity after the loading process. It then goes to 100% and stays there, locking up all functions. Meanwhile, my Linux computer at home has chugged away for about 15 years spitting out Seti and now Einstein work units like clockwork.

Win10 Pro 20H2
Boinc version 7.16.11 (x64)
ID: 103374 · Report as offensive
Profile Keith Myers
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 17 Nov 16
Posts: 863
United States
Message 103375 - Posted: 2 Mar 2021, 21:03:47 UTC - in response to Message 103374.  

Reinstall Windows. Doubt there is anything wrong with BOINC.
ID: 103375 · Report as offensive
Dr Who Fan
Avatar

Send message
Joined: 10 May 07
Posts: 1328
United States
Message 103376 - Posted: 2 Mar 2021, 21:10:22 UTC - in response to Message 103375.  

Reinstall Windows. Doubt there is anything wrong with BOINC.

Agree with slightly different idea... Reboot in safe mode and run a scandisk or chkdsk with automatic repair to make sure nothing is wrong with the hard drive.
ID: 103376 · Report as offensive
Profile Dave
Help desk expert

Send message
Joined: 28 Jun 10
Posts: 2515
United Kingdom
Message 103377 - Posted: 2 Mar 2021, 21:13:25 UTC

I would cut down the number of cores crunching, then use task manager to look at what disk activity there is. Like Dr Who Fan, I suspect something other than BOINC is causing the problem.

Just a thought. Check memory usage. Is there very heavy use of swap file?
ID: 103377 · Report as offensive
tbzep

Send message
Joined: 29 Jan 17
Posts: 9
United States
Message 103385 - Posted: 3 Mar 2021, 13:03:01 UTC
Last modified: 3 Mar 2021, 13:30:25 UTC

It is my work computer (govt owned) so I can't reinstall Windows. Memory usage looks normal. CPU usage drops a little. At the point where disk usage maxes out, task manager locks up as well.

I will play with CPU core numbers and see if that keeps it from locking up so I can get some data from task manager.

Edit: Just ran it with 3 cores (six threads) and noticed the memory usage doubled on each work unit as the disk usage climbed to 100%. Using 3 cores gave me an extra 20-30 seconds to watch task manager but it still locked up the system.

I didn't made any changes to the system or Boinc settings when it started. I did the update windows update after the problem began. I wondered if there may have been an antivirus update or a change pushed out by our sysadmins. I put an exception in the antivirus software so it would not scan Boinc operations the other day, but it had no affect.

Safe mode chkdisk is next.
ID: 103385 · Report as offensive
Profile Dave
Help desk expert

Send message
Joined: 28 Jun 10
Posts: 2515
United Kingdom
Message 103386 - Posted: 3 Mar 2021, 14:25:19 UTC

It would be useful to know what percentage free memory you have and what the total is before it locks up. Possibly even going down another couple of threads if it gives you enough time to get some figures. I know that on my Ryzen 5 CPDN tasks, a couple of browser windows and a couple of word processing documents open can at peak usage get me down to 1% of RAM free. My swap file is on an NVME SSD disk. I suspect that if it was on an older slower disk it would certainly slow things down a lot as the CPU waited to write/read data and also slow down disk access for other things. I have no idea what Einstein is like for memory usage though.
ID: 103386 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5077
United Kingdom
Message 103387 - Posted: 3 Mar 2021, 14:34:41 UTC - in response to Message 103386.  

In that case, it would also help to consider which of the many sectors of scientific research that Einstein comprises were active at the time. Their Gravity Wave search, in particular, relies on having something like 48 downloaded data files, some 150MB - 200MB of raw data, present while the task is running. I don't have figures for the effect this has on memory usage by the application, but I suspect it can't be small.
ID: 103387 · Report as offensive
Dr Who Fan
Avatar

Send message
Joined: 10 May 07
Posts: 1328
United States
Message 103388 - Posted: 3 Mar 2021, 14:52:56 UTC - in response to Message 103385.  

It is my work computer (govt owned)

Did you get permission from your employer to run BOINC on the computer??
+++++++++++++++++++
If Yes to the above... Then we have several things to investigate:

* Are the BOINC folders and subfolders (program & data) and files excluded from the active scan/monitor of the antivirus software?
>>> [If not, excluded from active scan]

* How much physical RAM does the PC have?

* How much of the RAM is available/free before loading BOINC?

* How large is the SWAP FILE; is the SWAP FILE set to a fixed or variable amount?
ID: 103388 · Report as offensive
tbzep

Send message
Joined: 29 Jan 17
Posts: 9
United States
Message 103391 - Posted: 3 Mar 2021, 15:45:50 UTC - in response to Message 103388.  

Yes, I have permission to run it. I started running Seti@Home from work in 2001.

I noticed task manager lagged behind events that were happening so I put Rainmeter on the desktop. I also timed the events. At approximately 4:20 after starting computations with 3 of 4 cores (and GPU), I saw hard disk and memory both max out. Taskmanager was locking up before I saw changes in memory.

I excluded Boinc folders from AVG antivirus last week.

8gb ram total. 2.1gb is used at idle (with browser open). 3.1gb is used for the first ~4 minutes of running computations with 3 of 4 cores. I normally run all 4 cores, but there is still plenty of headroom normally.

HD limit is set at 50gb in Boinc, which is what it has always been. I have 357gb of free space on the drive.

Pagefile has always been set variable. Right now it is at 13,824mb with the usual stuff opened like the browser, Boinc (at idle), etc.
ID: 103391 · Report as offensive
tbzep

Send message
Joined: 29 Jan 17
Posts: 9
United States
Message 103392 - Posted: 3 Mar 2021, 15:57:49 UTC - in response to Message 103387.  

In that case, it would also help to consider which of the many sectors of scientific research that Einstein comprises were active at the time. Their Gravity Wave search, in particular, relies on having something like 48 downloaded data files, some 150MB - 200MB of raw data, present while the task is running. I don't have figures for the effect this has on memory usage by the application, but I suspect it can't be small.

I ran a split of Seti and Einstein for a year or two before Seti shut down. I've been running Einstein on this computer ever since. This issue started week before last. I've always run all four cores and GPU with this computer and allowed 100% CPU and memory use when the computer isn't in use. Suspend is set at 25%. A shake of the mouse has always suspended it within a second or two. With it set to "run always", I have always been able to browse, and do minor stuff with minor lagging. Unless Einstein work units have changed in the last couple of weeks, it shouldn't be an issue. However, I did purge all work units and started new ones last week to see if there was a bad one causing havoc.
ID: 103392 · Report as offensive
Profile Dave
Help desk expert

Send message
Joined: 28 Jun 10
Posts: 2515
United Kingdom
Message 103394 - Posted: 3 Mar 2021, 16:07:39 UTC - in response to Message 103391.  

Yes, I have permission to run it. I started running Seti@Home from work in 2001.

I noticed task manager lagged behind events that were happening so I put Rainmeter on the desktop. I also timed the events. At approximately 4:20 after starting computations with 3 of 4 cores (and GPU), I saw hard disk and memory both max out. Taskmanager was locking up before I saw changes in memory.

I excluded Boinc folders from AVG antivirus last week.

8gb ram total. 2.1gb is used at idle (with browser open). 3.1gb is used for the first ~4 minutes of running computations with 3 of 4 cores. I normally run all 4 cores, but there is still plenty of headroom normally.

HD limit is set at 50gb in Boinc, which is what it has always been. I have 357gb of free space on the drive.

Pagefile has always been set variable. Right now it is at 13,824mb with the usual stuff opened like the browser, Boinc (at idle), etc.


I do wonder if you have recently been sent tasks that are much heavier on memory usage. 8GB really isn't very much. I found this on the Einstein Forums

Who would have thought?
I started running out of memory when I switched to nearly all E@H CPU tasks. I had to up the memory on the AMD 3950x box to 32GB again.
ID: 103394 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5077
United Kingdom
Message 103395 - Posted: 3 Mar 2021, 16:17:19 UTC - in response to Message 103392.  

The point I was trying to make was that Einstein has more than one type of work, and they change over time. Looking at their applications page today, they have two applications for CPUs at the moment:

Gravitational Wave search O2 Multi-Directional (O2MD1) - started 7 Dec 2020
Gamma-ray pulsar search #5 (FGRP5) - started 1 Aug 2017

Given the dates, it's possible that there's a particular problem with the O2MD1 app on your hardware. Have you discussed this on the Einstein message board yet?

(BTW, you'll see that the Windows version of O2MD1 was revised on 1 Feb 2021. You can read about that here - it won't affect your problem).
ID: 103395 · Report as offensive
tbzep

Send message
Joined: 29 Jan 17
Posts: 9
United States
Message 103397 - Posted: 3 Mar 2021, 16:33:41 UTC - in response to Message 103395.  
Last modified: 3 Mar 2021, 16:39:11 UTC

Understood, and you may be on to something. I was just pointing out that I've run Einstein a long time and see the different work unit names all the time. I didn't think it was the problem since it runs fine for over 4 minutes before going nuts.



However, I just started it using a single core (two work units) and closely watched the memory and disk usage. At about 4.5 minutes the memory usage doubled, but well within the 8gb limit. The disk stayed at zero. It does seem to be a page swapping issue, but I don't understand why it runs fine for more than four minutes and then doubles the memory usage.

Edit: actually 3 units running, 2 on the single core and 1 on the GPU.
ID: 103397 · Report as offensive
Richard Haselgrove
Volunteer tester
Help desk expert

Send message
Joined: 5 Oct 06
Posts: 5077
United Kingdom
Message 103398 - Posted: 3 Mar 2021, 16:52:42 UTC - in response to Message 103397.  

Edit: actually 3 units running, 2 on the single core and 1 on the GPU.
Ah. You hadn't mentioned a GPU. That brings O2MDF and FGRPB1G into play as well, with a wider range of dates.

My understanding is that for the tasks using the large data files - the Gravity Wave tasks - the first action is to read and unpack the raw data. Possibly, the portions related to the parameters of the current task have to be extracted. But only after the data pre-processing is complete, can the mathematical processing begin. Only the Einstein team would be able to guide you through all that.
ID: 103398 · Report as offensive
tbzep

Send message
Joined: 29 Jan 17
Posts: 9
United States
Message 103399 - Posted: 3 Mar 2021, 17:04:07 UTC - in response to Message 103398.  

Edit: actually 3 units running, 2 on the single core and 1 on the GPU.
Ah. You hadn't mentioned a GPU. That brings O2MDF and FGRPB1G into play as well, with a wider range of dates.

My understanding is that for the tasks using the large data files - the Gravity Wave tasks - the first action is to read and unpack the raw data. Possibly, the portions related to the parameters of the current task have to be extracted. But only after the data pre-processing is complete, can the mathematical processing begin. Only the Einstein team would be able to guide you through all that.


Thanks, Richard and everyone. I've posted the info on the Einstein page to see what they say. I may just have to stop running Einstein on this computer or keep running it on a single core.

Dr. Who Fan, have you often dreamed of using the Tardis to go back in time to make sure Alou caught that foul ball? :)
ID: 103399 · Report as offensive
Dr Who Fan
Avatar

Send message
Joined: 10 May 07
Posts: 1328
United States
Message 103404 - Posted: 3 Mar 2021, 20:16:13 UTC - in response to Message 103399.  

@tbzep
Dr. Who Fan, have you often dreamed of using the Tardis to go back in time to make sure Alou caught that foul ball? :)

As a die-hard Blue-blooded Cubs fan Yes I do. I can never forgive "the Bartman" for interfering with the play.
ID: 103404 · Report as offensive

Message boards : BOINC Manager : Boinc 100% disk activity, locks up

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.