climateprediction.net home page
How to Prevent OpenIFS Download

How to Prevent OpenIFS Download

Message boards : Number crunching : How to Prevent OpenIFS Download
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
AndreyOR

Send message
Joined: 12 Apr 21
Posts: 247
Credit: 12,074,507
RAC: 2,078
Message 68075 - Posted: 27 Jan 2023, 7:51:56 UTC - in response to Message 68068.  

i have a 8GB machine that runs a single task at a time with 100% success rate.
i have a 16GB machines that runs 2 tasks concurrently with 100% success rate.
i believe any issue is not about absolute memory available, but how those machines run.

I agree with your last statement but only partially. Partially because Glenn has said a number of times that pushing the RAM too much is a big reason for task failures and that he doesn't think 8GB RAM PCs should attempt these tasks. That probably means that you're in the minority with having been able to run almost 30 tasks error free so far on an 8GB machine. That's good for you that you're able to put that machine to use here.

Looking at your 16GB PC, the error rate is ~8% (6/78). That's not bad, under 5% is the stated goal though. I have an older 16GB machine that I started with 2 concurrent tasks and was getting errors. They pretty much disappeared when i went down to 1.
ID: 68075 · Report as offensive     Reply Quote
Vato

Send message
Joined: 4 Oct 19
Posts: 13
Credit: 7,353,923
RAC: 3,385
Message 68080 - Posted: 27 Jan 2023, 13:10:35 UTC - in response to Message 68075.  

i had not noticed those compute errors - apologies for the incorrect info and thank you for pointing it out.
i am pretty certain that the times when they occurred i was doing something else on that machine, which i ought to stop doing.
ID: 68080 · Report as offensive     Reply Quote
Glenn Carver

Send message
Joined: 29 Oct 17
Posts: 847
Credit: 13,960,536
RAC: 6,962
Message 68083 - Posted: 27 Jan 2023, 15:37:26 UTC - in response to Message 68080.  
Last modified: 27 Jan 2023, 15:37:48 UTC

i had not noticed those compute errors - apologies for the incorrect info and thank you for pointing it out.
i am pretty certain that the times when they occurred i was doing something else on that machine, which i ought to stop doing.
I had a quick look. Those error code 9 fails indicate some memory corruption. I have an impression running less tasks, or doing less on the machine helps but I don't have any numbers to prove it. I wouldn't want anyone to not use their computers to avoid any random error like this. I've eliminated the ones I can find it our code but am stuck on the ones coming from the boinc client. We're looking at it but I can't say when it'll be fixed.
ID: 68083 · Report as offensive     Reply Quote
Yeti

Send message
Joined: 5 Aug 04
Posts: 172
Credit: 11,067,868
RAC: 28,961
Message 68094 - Posted: 27 Jan 2023, 22:53:37 UTC

I think, one very important setting regarding memory is:

Memory
When computer is in use, use at most 100 %
When computer is not in use, use at most 100 %

It hasn't to be 100%, but both should use the same size.

If these differ from each other and you start using your machine, BOINC has to free up the memory to the lower limit and this seems to be very risky with CPDN-Tasks / OpenIFS
Supporting BOINC, a great concept !
ID: 68094 · Report as offensive     Reply Quote
gemini8

Send message
Joined: 4 Dec 15
Posts: 52
Credit: 2,182,959
RAC: 836
Message 68107 - Posted: 29 Jan 2023, 17:18:12 UTC - in response to Message 68094.  

Memory
When computer is in use, use at most 100 %
When computer is not in use, use at most 100 %

It hasn't to be 100%, but both should use the same size.

Ran my Boinc like this for years, because it seems sensible.
Also set use of the Swap to 1%.
- - - - - - - - - -
Greetings, Jens
ID: 68107 · Report as offensive     Reply Quote
Yeti

Send message
Joined: 5 Aug 04
Posts: 172
Credit: 11,067,868
RAC: 28,961
Message 68121 - Posted: 30 Jan 2023, 13:57:40 UTC - in response to Message 68107.  

Also set use of the Swap to 1%.
Not shure if this is really a good idea.

As long, as I have enough free Main-Memory, swap isn't needed and all is okay.

If you reduce swap-usage to 1% and your memory gets low, there is not much, what BOINC can do, so it will set a task to sleep or error something out.

If you allow swapping there is a chance that the running WU may survive and this will be a better chance as with blocked swapping.

Shure, swapping is no good for a lot of projects, but if BOINC has to think about swapping you/we have already made a big mistake
Supporting BOINC, a great concept !
ID: 68121 · Report as offensive     Reply Quote
gemini8

Send message
Joined: 4 Dec 15
Posts: 52
Credit: 2,182,959
RAC: 836
Message 68180 - Posted: 2 Feb 2023, 8:10:10 UTC

Actually I don't really know how this works.
Does Boinc use just this bit of the already existing swap the system has, or does Boinc use additional disk space?

I'm running Linux which has a very small footprint, so if the swap gets swamped it should be other things than Linux itself.
On a cruncher which does no other stuff on the side, swap should hardly be used, especially with my settings.
So if it is used nonetheless, Boinc doesn't behave correctly, does it?

Actually I could see that with CPDN tasks remaining in memory even when they are put aside the system could offload them into swap until they're active again.
- - - - - - - - - -
Greetings, Jens
ID: 68180 · Report as offensive     Reply Quote
Previous · 1 · 2

Message boards : Number crunching : How to Prevent OpenIFS Download

©2024 climateprediction.net