Some Rosetta files do not progress anymore

Questions and Answers : Windows : Some Rosetta files do not progress anymore

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
beguwe1

Send message
Joined: 7 Mar 10
Posts: 1
Credit: 178,223
RAC: 0
Message 78943 - Posted: 18 Oct 2015, 10:11:52 UTC

Hi,
I am having the following issue sometimes on more than one PC:
- some Rosetta files progress till a certain %
- then the processing time is still increasing, but the progress % indicator stays as-is.
- so no progress anymore on the file.
-if the 'estimated time to complete' is at zero, there is a dash (-) placed in that column.
- rebooting the PC does not help, only 'work aboard' helps.

==> anybody knows the root cause of this issue?
(i have BOINC version 7.6.9 (Windows x64)

Best Regards,
/Guy
Belgium
ID: 78943 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
dsalt2

Send message
Joined: 30 Apr 07
Posts: 1
Credit: 2,803,452
RAC: 2,404
Message 80102 - Posted: 19 May 2016, 9:44:55 UTC

I have exactly the same problem!
Tasks get to a certain percentage then stop going up in % but the run-time carries on increasing for ever and never stops.
Rebooting doesn't work, I end up having to abort the tasks.
HELP!

Rosetta 3.73
Boinc 7.6.22 (x64)
Windows 10 x64
ID: 80102 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 80104 - Posted: 19 May 2016, 13:19:21 UTC
Last modified: 19 May 2016, 13:20:45 UTC

If rebooting the machine does not get things going well again, then it is not the problem I often see where BOINC Manager does not actually give the thread CPU time (even though it shows the task as "running"). So this may be an issue with specific tasks. A Link to the task and the task name might be helpful. And the Number Crunching board has many more readers than the Q&A boards, so I'd suggest posting the details there.

Is it possible that the BOINC Manager ran out of disk space available?
Rosetta Moderator: Mod.Sense
ID: 80104 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mark

Send message
Joined: 11 Oct 16
Posts: 8
Credit: 363,827
RAC: 0
Message 80744 - Posted: 16 Oct 2016, 13:08:49 UTC

Bump.

Stuck tasks: The recommended first-aid is to hit "Reset project," which removes all pending tasks. After that, a new batch of tasks arrives; 6 of them begin to run; one finishes within a few minutes and uploads results to headquarters; the others get stuck.

Stuck tasks load gigabytes of data into RAM, where it remains even when the tasks are suspended. Screen grab from Windows Task Manager, identical for running and suspended tasks:
https://postimg.org/image/fherq5zcp/

Copious unused disk space and RAM space remains.

-----

Windows 10 Home 64-bit, x-64-based processor
Rosetta Mini 3.73
BOINC Manager 7.6.22 (x64)
ID: 80744 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 80745 - Posted: 16 Oct 2016, 14:43:38 UTC
Last modified: 16 Oct 2016, 14:47:02 UTC

Your screenshot shows 0% CPU going to the tasks. The BOINC Manager determines what gets CPU and what doesn't. The memory they consume is normal. And if the tasks are not running, then Windows will swap them out of memory. But if your BOINC settings do not allow BOINC to use more memory, that could prevent tasks from progressing.

There are a number of possibilities for why tasks might not be getting CPU. If the screenshot doesn't show all of the BOINC work units, perhaps other work units are running and using all of the memory BOINC is allowed to use. Also, if there are other, non-BOINC, tasks keeping the machine busy, BOINC Manager has a threshold of CPU % busy where once your machine gets too busy, it steps out of the way and suspends BOINC activity until the machine is less busy.

I'd suggest completely ending BOINC Manager ("exit", not "close"), and restarting. Then check the BOINC Manager events log and the tasks display in the advanced view to see if there are indications of why tasks are being suspended. Also, if you could explain more about your BOINC settings for the computing preferences, that would help put things in perspective.
Rosetta Moderator: Mod.Sense
ID: 80745 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mark

Send message
Joined: 11 Oct 16
Posts: 8
Credit: 363,827
RAC: 0
Message 80746 - Posted: 16 Oct 2016, 19:49:44 UTC

Thanks for the reply.

The screenshot shows the typical situation when BOINC Manager says that 6 tasks are running. The CPU is at 0% overall, not just for Rosetta tasks, so plenty of CPU is available. RAM shows as 39% in use -- that's 61% or 9.6 GB free, so plenty of RAM is available. And 105 GB of disk space is available.

Exiting and restarting BOINC Manager doesn't help.

Rebooting the computer doesn't help.

After an exit-restart, the Event Log reports "Resuming computation" and "Resuming network activity" and gives no indication of a problem.

SETTINGS
Use at most 75% of the CPUs
Use at most 75% of CPU time
Suspend when computer is in use
'In use" means...3 minutes
Suspend when non-BOINC CPU usage is above 25%
Network: default settings
Disk: Leave at least 0.1 GB free
Disk: Use no more than 90% of total
Memory: When computer is in use, use at most 50% [not applicable]
Memory: When computer is not in use use, use at most 90%
Memory: Page/swap file: use at most 75%
Daily schedules: default settings (blank)
ID: 80746 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Sid Celery

Send message
Joined: 11 Feb 08
Posts: 2125
Credit: 41,245,383
RAC: 9,571
Message 80749 - Posted: 17 Oct 2016, 22:43:56 UTC - in response to Message 80746.  

Thanks for the reply.

The screenshot shows the typical situation when BOINC Manager says that 6 tasks are running. The CPU is at 0% overall, not just for Rosetta tasks, so plenty of CPU is available. RAM shows as 39% in use -- that's 61% or 9.6 GB free, so plenty of RAM is available. And 105 GB of disk space is available.

Exiting and restarting BOINC Manager doesn't help.

Rebooting the computer doesn't help.

After an exit-restart, the Event Log reports "Resuming computation" and "Resuming network activity" and gives no indication of a problem.

SETTINGS
Use at most 75% of the CPUs
Use at most 75% of CPU time
Suspend when computer is in use
'In use" means...3 minutes
Suspend when non-BOINC CPU usage is above 25%
Network: default settings
Disk: Leave at least 0.1 GB free
Disk: Use no more than 90% of total
Memory: When computer is in use, use at most 50% [not applicable]
Memory: When computer is not in use use, use at most 90%
Memory: Page/swap file: use at most 75%
Daily schedules: default settings (blank)

Try these changes:

Use at most 100% of the CPUs
Use at most 100% of CPU time
Memory: When computer is in use, use at most 65%
Leave non-GPU tasks in memory when suspended - Tick

Did you say how much RAM you have?

ID: 80749 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 80750 - Posted: 18 Oct 2016, 12:07:19 UTC - in response to Message 80749.  
Last modified: 18 Oct 2016, 12:08:40 UTC

Did you say how much RAM you have?


They have 8 CPUs (with max 75% of CPUs) and 16GB.

Is there a known bug in BOINC Manager? Seems it should always have at least one task running if memory permits (and it does).
Rosetta Moderator: Mod.Sense
ID: 80750 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mark

Send message
Joined: 11 Oct 16
Posts: 8
Credit: 363,827
RAC: 0
Message 80751 - Posted: 18 Oct 2016, 19:17:32 UTC - in response to Message 80750.  
Last modified: 18 Oct 2016, 19:18:28 UTC

Did you say how much RAM you have?


They have 8 CPUs (with max 75% of CPUs) and 16GB.

Is there a known bug in BOINC Manager? Seems it should always have at least one task running if memory permits (and it does).


Correct.

Re. your list in an earlier post of circumstances that cause BOINC Manager to deny CPU to tasks: My expectation is that if BOINC Manager purposely suspends activity, for the reasons you suggest or any other, then the UI ought to reflect this state of affairs. The "Status" column should say something other than "Running," the "Elapsed" column should stop ticking off the seconds, and the event log should give some hint. None of that happened, I think, for the users posting in this thread.

I am thrilled and grateful that you and Sid are giving this matter some attention. Before posting here, I tried raising the issue in 3 other locations (BOINC, MilkyWay, email support) and received no response. So thanks.
ID: 80751 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mark

Send message
Joined: 11 Oct 16
Posts: 8
Credit: 363,827
RAC: 0
Message 80752 - Posted: 18 Oct 2016, 19:20:31 UTC - in response to Message 80749.  


Try these changes:

Use at most 100% of the CPUs
Use at most 100% of CPU time
Memory: When computer is in use, use at most 65%
Leave non-GPU tasks in memory when suspended - Tick

Did you say how much RAM you have?


RAM: 16.0 GB

All right. I uninstalled BOINC, cleaned up, downloaded a fresh installer, installed BOINC, and (sceptically) applied your suggested settings. Also changed "Suspend when non-BOINC CPU usage is above" to 40% from 25%. (The original settings were copied from a 5-year-old, much less powerful Windows PC, which was running BOINC to perfection, though glacially.)

So far, so good. BOINC and Rosetta appear to be operating as expected.

----------

Rosetta Mini 3.73
BOINC Manager 7.6.22 (x64)
---
Windows 10 Home 64-bit, x-64-based processor, English (Canada)
--
ASUS Desktop PC M32CD Series
Intel(R) Core(TM) i7-6700 CPU @ 3.40 GHz
Skylake 14nm Technology
Motherboard: ASUSTeK COMPUTER INC. M32CD_A_F_K20CD_K31CD (LGA1151)
RAM: 16.00 GB
GPU: AMD Radeon R9 370
--
Windows Defender
Windows Firewall
WinPatrol PLUS
ID: 80752 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Sid Celery

Send message
Joined: 11 Feb 08
Posts: 2125
Credit: 41,245,383
RAC: 9,571
Message 80753 - Posted: 19 Oct 2016, 3:45:16 UTC - in response to Message 80750.  

Is there a known bug in BOINC Manager?

Don't even go there... ;)
ID: 80753 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Sid Celery

Send message
Joined: 11 Feb 08
Posts: 2125
Credit: 41,245,383
RAC: 9,571
Message 80754 - Posted: 19 Oct 2016, 3:56:00 UTC - in response to Message 80752.  
Last modified: 19 Oct 2016, 4:02:15 UTC

Try these changes:

Use at most 100% of the CPUs
Use at most 100% of CPU time
Memory: When computer is in use, use at most 65%
Leave non-GPU tasks in memory when suspended - Tick

Did you say how much RAM you have?

RAM: 16.0 GB

I also have an 8-core machine with 16Gb ram, but an AMD. By and large everything should be the same between us.

Also changed "Suspend when non-BOINC CPU usage is above" to 40% from 25%. (The original settings were copied from a 5-year-old, much less powerful Windows PC, which was running BOINC to perfection, though glacially.)

So far, so good. BOINC and Rosetta appear to be operating as expected.

I didn't notice this first time round. While the other changes are worth doing, this seems more significant. Given that all Rosetta tasks run at low-priority, and all non-Boinc tasks run at a higher setting (Normal) this setting seems redundant. It might've been necessary with an underpowered PC, but imo serves no purpose with the power of your current machine. All non-Boinc processing has a higher priority from the first 1%. Suspending tasks altogether when it gets above a certain point seems like overkill to me.

I have this option unticked and don't notice a problem. Why waste time even checking, given the above? See if you notice any difference (though your change may be sufficient in itself).
ID: 80754 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 80755 - Posted: 19 Oct 2016, 4:12:33 UTC - in response to Message 80751.  

My expectation is that if BOINC Manager purposely suspends activity, for the reasons you suggest or any other, then the UI ought to reflect this state of affairs. The "Status" column should say something other than "Running," the "Elapsed" column should stop ticking off the seconds, and the event log should give some hint. None of that happened, I think, for the users posting in this thread.


Agreed. I believe when the problem occurs, the elapsed time increases, but in the properties of the tasks, you'll see that the CPU time is NOT increasing. This also results in outrageous "estimated" time to completion (since this is just computed from the fraction done and the elapsed time).

If Sid's suggested settings continue to work for you (and you have time and interest to tinker further), I'd suggest changing settings one-at-a-time until you see the zombie tasks again. The most significant to your environment I should think would be the 75% of the CPUs, thus leaving you two CPUs always available for non-BOINC activity. You must have desired that setting to keep a reserve for some other activity on the machine.

Rosetta Moderator: Mod.Sense
ID: 80755 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mark

Send message
Joined: 11 Oct 16
Posts: 8
Credit: 363,827
RAC: 0
Message 80757 - Posted: 19 Oct 2016, 20:50:29 UTC

...I'd suggest changing settings one-at-a-time until you see the zombie tasks again.


Here are the results. Each of the options referred to in this thread was reverted, one at a time, to its previous setting. The culprit turned out to be "Use at most 75% of CPU time." At that setting, all heck broke loose. CPU utilization showed wild swings, overall and for each Rosetta task. After a minute, one of those tasks went to 0% and turned zombie as described above. Life returned when the setting was changed back to 100%, but only after BOINC Manager was closed and the computer rebooted.
ID: 80757 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 80758 - Posted: 19 Oct 2016, 21:11:14 UTC

So you have two CPU settings, 75% of the number of CPUs, and 75% of total CPU power. So were you able to get the 75% of CPUs (i.e. 6 CPUs) running at 100%?
Rosetta Moderator: Mod.Sense
ID: 80758 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mark

Send message
Joined: 11 Oct 16
Posts: 8
Credit: 363,827
RAC: 0
Message 80759 - Posted: 19 Oct 2016, 22:46:38 UTC - in response to Message 80758.  

So you have two CPU settings, 75% of the number of CPUs, and 75% of total CPU power. So were you able to get the 75% of CPUs (i.e. 6 CPUs) running at 100%?

On my system, when BOINC is configured for 75% of CPUs and the evidently mandatory 100% of CPU power, BOINC projects receive a steady 75% of CPU and each task gets about 12.5% (75/6, 100/8). My system can't handle the "75% of CPU power" setting, as described above.
ID: 80759 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Sid Celery

Send message
Joined: 11 Feb 08
Posts: 2125
Credit: 41,245,383
RAC: 9,571
Message 80760 - Posted: 20 Oct 2016, 2:13:30 UTC - in response to Message 80757.  

...I'd suggest changing settings one-at-a-time until you see the zombie tasks again.

Here are the results. Each of the options referred to in this thread was reverted, one at a time, to its previous setting. The culprit turned out to be "Use at most 75% of CPU time." At that setting, all heck broke loose. CPU utilization showed wild swings, overall and for each Rosetta task. After a minute, one of those tasks went to 0% and turned zombie as described above. Life returned when the setting was changed back to 100%, but only after BOINC Manager was closed and the computer rebooted.

A long time ago, using a very different machine and a much older Boinc, I set a reduced %age of CPU time and noticed that what actually happened was 100% of CPU time was used for (say) 75% of the time and 0% for the remainder of the time, switching multiple times per second. I always knew it was a weird setting, though I didn't know it would do what you've reported. Interesting to note.

So you're happy with how things are going? Good to hear, if so.
ID: 80760 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mark

Send message
Joined: 11 Oct 16
Posts: 8
Credit: 363,827
RAC: 0
Message 80763 - Posted: 20 Oct 2016, 12:56:48 UTC

A long time ago, using a very different machine and a much older Boinc, I set a reduced %age of CPU time and noticed that what actually happened was 100% of CPU time was used for (say) 75% of the time and 0% for the remainder of the time, switching multiple times per second. I always knew it was a weird setting, though I didn't know it would do what you've reported. Interesting to note.

So you're happy with how things are going? Good to hear, if so.


Ah, that may explain what I was seeing in the Windows Task Manager at the 75% setting.

Happy? Yes. Although if a reduced CPU time setting ever becomes workable for me, I certainly want to take advantage -- to lower CPU temperature.
ID: 80763 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 80764 - Posted: 20 Oct 2016, 15:36:59 UTC

The only means BOINC Manager has of implementing "75% of CPU time" is to go 100% for a period of time and then go zero for a period of time to average out to 75%. And so far as CPU temps go, that's all you need. The temps average out the same way the CPU % does. Computers don't just run at 75%. They're either on or off. Even when there is nothing to do, it has to spin the extra cycles in the idle task. The gotta go somewhere.

When using 6 out of the 8 CPUs (i.e. 75% of the processors rather than of the CPU) I'm guessing the tasks settle in to specific CPUs and don't leave. In other words you have 6 CPUs running 100% of the time, and two that are only supporting operating system overhead and other non-BOINC tasks. The good news is most systems are properly designed to deliver sufficient airflow to keep the CPUs are a healthy temp.
Rosetta Moderator: Mod.Sense
ID: 80764 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Sid Celery

Send message
Joined: 11 Feb 08
Posts: 2125
Credit: 41,245,383
RAC: 9,571
Message 80765 - Posted: 21 Oct 2016, 3:05:36 UTC - in response to Message 80763.  

A long time ago, using a very different machine and a much older Boinc, I set a reduced %age of CPU time and noticed that what actually happened was 100% of CPU time was used for (say) 75% of the time and 0% for the remainder of the time, switching multiple times per second. I always knew it was a weird setting, though I didn't know it would do what you've reported. Interesting to note.

So you're happy with how things are going? Good to hear, if so.

Ah, that may explain what I was seeing in the Windows Task Manager at the 75% setting.

Happy? Yes. Although if a reduced CPU time setting ever becomes workable for me, I certainly want to take advantage -- to lower CPU temperature.

I'm not sure I ever saw it in the Task Manager. I used to use a sidebar gadget under Vista at one time that revealed it.

Temperatures are certainly an issue. I'm overclocking and use HWInfo64 to keep an eye on that. If the 75% of CPU's setting (for 6 of 8 cores) worked ok for you (others report it working fine too) wouldn't that knock the overall temperature down? I don't know as I've never tried it but definitely worth experimenting with for a few days. A good halfway-house setting and more like what you were setting out to do in the first place it seems to me.
ID: 80765 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
1 · 2 · Next

Questions and Answers : Windows : Some Rosetta files do not progress anymore



©2024 University of Washington
https://www.bakerlab.org