how long can "long tasks" be?

Message boards : Number crunching : how long can "long tasks" be?
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
doug

Send message
Joined: 9 May 22
Posts: 2
Credit: 1,350,578
RAC: 123
Message 1809 - Posted: 15 Jan 2023, 15:56:07 UTC

Hi all,

How long can the tasks with "long tasks 2.00" in the name take? They've been reasonably fast for me in the past (hours, not days), but now I have a couple that have been running for approx. 2.5 and 1+ days, and the first one says it has 5+ days left and the second one says it has 8.5 days left? BOINCTasks shows these two are running at 98% and 97.6% CPU%, so they at least haven't died.

Does it make sense that I now have tasks that would be this long? If not, should I just kill them?

Thanks.

Doug
ID: 1809 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
marmot

Send message
Joined: 10 Dec 20
Posts: 24
Credit: 10,767,590
RAC: 0
Message 1811 - Posted: 15 Jan 2023, 16:52:49 UTC - in response to Message 1809.  
Last modified: 15 Jan 2023, 16:54:50 UTC

Something might be wrong with those tasks.
Use your Process Manager and see if they actually are using CPU.

I'd kill those.

Or do what the admin says in this message:
https://www.sidock.si/sidock/forum_thread.php?id=224&postid=1787#1787
ID: 1811 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Bryn Mawr

Send message
Joined: 16 Aug 21
Posts: 40
Credit: 17,533,635
RAC: 28,979
Message 1812 - Posted: 15 Jan 2023, 17:16:57 UTC - in response to Message 1811.  

Something might be wrong with those tasks.
Use your Process Manager and see if they actually are using CPU.

I'd kill those.

Or do what the admin says in this message:
https://www.sidock.si/sidock/forum_thread.php?id=224&postid=1787#1787


They are most likely using 100% CPU.

In computing options set leave on memory off, then read config files.
Suspend the task(s) then resume
Set leave in memory back on then read config files.

That will force the task(s) to restart at the last checkpoint and all should be ok.
ID: 1812 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
arcturus

Send message
Joined: 27 Nov 22
Posts: 20
Credit: 4,456,821
RAC: 23,465
Message 1816 - Posted: 15 Jan 2023, 18:13:17 UTC

The circumstances that led to the unit hanging hasn't changed so I just abort them to avoid the possibility the same thing happens again to the same task.
ID: 1816 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Bryn Mawr

Send message
Joined: 16 Aug 21
Posts: 40
Credit: 17,533,635
RAC: 28,979
Message 1818 - Posted: 15 Jan 2023, 23:19:34 UTC - in response to Message 1816.  

The circumstances that led to the unit hanging hasn't changed so I just abort them to avoid the possibility the same thing happens again to the same task.


No, the circumstances that led to the unit hanging are intermittent, by going back to the last checkpoint the likelihood of it happening again is the same as the likelihood of it happening in the first place.
ID: 1818 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
arcturus

Send message
Joined: 27 Nov 22
Posts: 20
Credit: 4,456,821
RAC: 23,465
Message 1819 - Posted: 15 Jan 2023, 23:57:22 UTC - in response to Message 1818.  
Last modified: 16 Jan 2023, 0:00:29 UTC

The circumstances that led to the unit hanging hasn't changed so I just abort them to avoid the possibility the same thing happens again to the same task.


No, the circumstances that led to the unit hanging are intermittent, by going back to the last checkpoint the likelihood of it happening again is the same as the likelihood of it happening in the first place.


In other words, the chance in intermittent. I'll rather avoid it completely and await the determination of the developer.
ID: 1819 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
marmot

Send message
Joined: 10 Dec 20
Posts: 24
Credit: 10,767,590
RAC: 0
Message 1867 - Posted: 18 Jan 2023, 3:05:32 UTC
Last modified: 18 Jan 2023, 3:10:18 UTC

All of my tasks are this long; which Bryn Mawr said was the intent.

My KabyLake 14nm Laptop shows 3d 4 hours left to complete which I'm not sure it can make by Jan 21.

The electric company put 4 hours of high cost peak periods that I have to pause the machines for daily. BOINC does not support 2 pauses per day. Only through task scheduler can boinccmd be used to dual pause boinc.

The laptop is low power so I pause it for one period but the servers got a dual BOINC install and so half the SiDock long run for 10 hours in daylight and the other half for 10 hours at night.

Not sure any of those can complete by the 21st running only 10 hours a day. They are server Xeons, but 4th gen and older.

They easily completed even the longest SRBase the last month on the 10 hour/10 hour dual plan.

SRBase provided us a multithread option we can setup in a app_config.xml to assure we'd meet the deadlines.

Is multithreaded planned here?
ID: 1867 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
hoarfrost
Volunteer moderator
Project administrator
Project developer

Send message
Joined: 11 Oct 20
Posts: 337
Credit: 25,678,026
RAC: 9,200
Message 1878 - Posted: 18 Jan 2023, 9:02:19 UTC - in response to Message 1867.  

My KabyLake 14nm Laptop shows 3d 4 hours left to complete which I'm not sure it can make by Jan 21.

4 days to deadline for this tasks are added on server side (if I rightly found this computer). :) Probably that for next bunches of tasks the deadline will be extended.
ID: 1878 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
der_Day

Send message
Joined: 9 Feb 21
Posts: 7
Credit: 7,473,030
RAC: 2,438
Message 1884 - Posted: 18 Jan 2023, 18:09:40 UTC - in response to Message 1878.  

would be very kind, if you can also extend the deadline for this 2 WUs

https://www.sidock.si/sidock/workunit.php?wuid=49613180
https://www.sidock.si/sidock/workunit.php?wuid=49610715
ID: 1884 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
hoarfrost
Volunteer moderator
Project administrator
Project developer

Send message
Joined: 11 Oct 20
Posts: 337
Credit: 25,678,026
RAC: 9,200
Message 1885 - Posted: 19 Jan 2023, 8:20:44 UTC - in response to Message 1884.  

would be very kind, if you can also extend the deadline for this 2 WUs

https://www.sidock.si/sidock/workunit.php?wuid=49613180
https://www.sidock.si/sidock/workunit.php?wuid=49610715

Added. :)
ID: 1885 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Ralph Little

Send message
Joined: 21 May 22
Posts: 1
Credit: 2,183,514
RAC: 0
Message 1888 - Posted: 19 Jan 2023, 15:53:43 UTC

I am also seeing tasks that are estimated for 8+ days for a single thread at 100% CPU on a pretty fast machine. That's not really acceptable since I share my compute resources between a few different projects. I will suspend sidock on my systems until this is sorted out.
ID: 1888 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
swiftmallard

Send message
Joined: 9 Apr 22
Posts: 3
Credit: 1,249,136
RAC: 0
Message 1889 - Posted: 19 Jan 2023, 16:20:27 UTC - in response to Message 1867.  

Is multithreaded planned here?


Multithreaded tasks would be wonderful for the long WUs. I only ever crunch one project at a time so between multithreading and keeping a smaller cache, deadline issues would quickly become a thing of the past.
ID: 1889 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
der_Day

Send message
Joined: 9 Feb 21
Posts: 7
Credit: 7,473,030
RAC: 2,438
Message 1890 - Posted: 19 Jan 2023, 17:38:41 UTC - in response to Message 1885.  

thank you
ID: 1890 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
hoarfrost
Volunteer moderator
Project administrator
Project developer

Send message
Joined: 11 Oct 20
Posts: 337
Credit: 25,678,026
RAC: 9,200
Message 1891 - Posted: 19 Jan 2023, 19:31:52 UTC - in response to Message 1888.  

I am also seeing tasks that are estimated for 8+ days for a single thread at 100% CPU on a pretty fast machine.

Looks as very rough runtime estimation from BOINC. Processing of some ligands needs more time and different parts of workunit can consume very different time.
"My current record" ~30 hours on Ryzen 3900X with SMT, but not "days".
In any case - good luck! :)
ID: 1891 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
PMH_UK

Send message
Joined: 23 Dec 20
Posts: 20
Credit: 1,360,768
RAC: 0
Message 1892 - Posted: 19 Jan 2023, 19:51:10 UTC - in response to Message 1885.  

Please add 5 days to all tasks for userid=364
Most will finish before that, some on faster PCs in time, but slower PCs estimate up to 9 days run time.

Paul.
ID: 1892 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
hoarfrost
Volunteer moderator
Project administrator
Project developer

Send message
Joined: 11 Oct 20
Posts: 337
Credit: 25,678,026
RAC: 9,200
Message 1893 - Posted: 19 Jan 2023, 21:21:50 UTC - in response to Message 1892.  

Please add 5 days to all tasks for userid=364
Most will finish before that, some on faster PCs in time, but slower PCs estimate up to 9 days run time.

Added. :)
ID: 1893 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
marmot

Send message
Joined: 10 Dec 20
Posts: 24
Credit: 10,767,590
RAC: 0
Message 1902 - Posted: 20 Jan 2023, 15:45:36 UTC - in response to Message 1878.  
Last modified: 20 Jan 2023, 15:55:17 UTC

My KabyLake 14nm Laptop shows 3d 4 hours left to complete which I'm not sure it can make by Jan 21.

4 days to deadline for this tasks are added on server side (if I rightly found this computer). :) Probably that for next bunches of tasks the deadline will be extended.


That laptop was running an Einstein Intel GU task.
It lowered the CPU effective frequencies from 2400 to 1300 (which is way more severe a drain than I realized).
Stopped all new Einstein work and all but 2 are going to make the deadline wish I'd known about the ability to ask for deadline extensions:

would be very kind, if you can also extend the deadline for this 2 WUs

https://www.sidock.si/sidock/workunit.php?wuid=49613180
https://www.sidock.si/sidock/workunit.php?wuid=49610715

Added. :)


I could use another 2 days to complete the rest of these, unaborted tasks, on the one machine. I'm user 279.
Thankyou
ID: 1902 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
marmot

Send message
Joined: 10 Dec 20
Posts: 24
Credit: 10,767,590
RAC: 0
Message 1903 - Posted: 20 Jan 2023, 16:00:51 UTC - in response to Message 1892.  
Last modified: 20 Jan 2023, 16:01:13 UTC

but slower PCs estimate up to 9 days run time.

Paul.


My one machine had 20 WU over 7days left and the time left kept advancing 5 seconds every second.

Pausing/unpausing didn't help.
I gave up on them as they appeared to be hung.

Anyway, might as well d/l fresh WU's as they will have more realistic deadlines and the credit gained form a 9 day task won't be worth it.
ID: 1903 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Jon

Send message
Joined: 24 Nov 22
Posts: 1
Credit: 692,415
RAC: 0
Message 1904 - Posted: 20 Jan 2023, 17:15:03 UTC

Running on a 5900x. Before Long/Short tasks I used to crunch a unit in maybe under 3hrs. not they are averaging 1-2days with a couple running past the deadline with 4-6day estimates. Still crunching, but i've stopped new tasks for now.

can you add 4 days to these WU please:
https://www.sidock.si/sidock/workunit.php?wuid=49641040
https://www.sidock.si/sidock/workunit.php?wuid=49643250
ID: 1904 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
hoarfrost
Volunteer moderator
Project administrator
Project developer

Send message
Joined: 11 Oct 20
Posts: 337
Credit: 25,678,026
RAC: 9,200
Message 1905 - Posted: 20 Jan 2023, 18:39:37 UTC - in response to Message 1904.  

Hi! :)
Running on a 5900x. Before Long/Short tasks I used to crunch a unit in maybe under 3hrs. not they are averaging 1-2days with a couple running past the deadline with 4-6day estimates. Still crunching, but i've stopped new tasks for now.

can you add 4 days to these WU please:
https://www.sidock.si/sidock/workunit.php?wuid=49641040
https://www.sidock.si/sidock/workunit.php?wuid=49643250

Added. :) New tasks is really long. Most of them between 20 and 32 hours. But if you see tasks with estimation like "days" you can check it properties in BOINC Manager: "CPU time at last checkpoint" and "CPU time". If it values very differs (for 1 hour, for example) - task is hung and need a restart. (For example - stop BOINC and start it again). Also, in beginning of task it can be a rough estimation.

Good luck!
ID: 1905 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
1 · 2 · Next

Message boards : Number crunching : how long can "long tasks" be?

©2024 SiDock@home Team