Advanced search

Message boards : Number crunching : no ATMMLs downloaded while many tasks from other project running

Author Message
Erich56
Send message
Joined: 1 Jan 15
Posts: 1132
Credit: 10,382,797,676
RAC: 28,992,384
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwat
Message 61906 - Posted: 24 Oct 2024 | 7:21:49 UTC

On one of my hosts which is equipped with 2 CPUs Intel Xeon E5 2667 v4 (8 cores + 8 HT each), 128GB ramdisk + 128GB system RAM, ATMMLs stopped being downloaded after I had started 12 "Theory" tasks from LHC 2 days ago.
At this point, the free space on the ramdisk is 34GB (The setting for disc usage in Boinc is "leave 2GB free"), and free space on the system RAM is about 90GB (the setting for RAM in Boinc is "use max. 90%, a Theory task needs max. 1,5GB RAM).
So, there should be plenty of resources left for having an ATMML running (as it used to be the case all time long, until 2 days ago).

My question now is: does anyone know what other/extra resources requirements come with these ATMML tasks?
Or is there any other "trap" I am running into, without realizing?

P.S.: I am aware that the server status page is showing zero unsent tasks. Still, all my other hosts succeed to download a new ATMML within 1-2 hours after the previous one was uploaded. And this has been the case with this host here also - until 2 days ago.

KeithBriggs
Send message
Joined: 29 Aug 24
Posts: 10
Credit: 899,100,000
RAC: 13,469,212
Level
Glu
Scientific publications
wat
Message 61910 - Posted: 24 Oct 2024 | 13:16:08 UTC - in response to Message 61906.

No hints from the event log?

The only change I had to make about a month ago for some tasks was this: Disk: use at most 1000 GB

I think I tripled it.

My 2 cents.

Erich56
Send message
Joined: 1 Jan 15
Posts: 1132
Credit: 10,382,797,676
RAC: 28,992,384
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwat
Message 61911 - Posted: 24 Oct 2024 | 13:26:42 UTC - in response to Message 61910.

No hints from the event log?

The only change I had to make about a month ago for some tasks was this: Disk: use at most 1000 GB

I think I tripled it.

My 2 cents.

No hints in the event log. It only says "no tasks available for ATMML" (which is, of course true for some time; but, as I said, my other hosts could download tasks after short time once the previous one got finished).

What exectly was the reason for making the change you mention?

KeithBriggs
Send message
Joined: 29 Aug 24
Posts: 10
Credit: 899,100,000
RAC: 13,469,212
Level
Glu
Scientific publications
wat
Message 61913 - Posted: 24 Oct 2024 | 14:07:13 UTC - in response to Message 61911.

I either got a notice that there wasn't room or I kept trying things to get a task.

My latest task only takes up about ~100 MB but it seemed like the server wasn't getting good info about my system and wasn't letting tasks in. I don't have 1000 GB available!

Erich56
Send message
Joined: 1 Jan 15
Posts: 1132
Credit: 10,382,797,676
RAC: 28,992,384
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwat
Message 61914 - Posted: 24 Oct 2024 | 15:41:35 UTC - in response to Message 61913.

I either got a notice that there wasn't room or I kept trying things to get a task.

My latest task only takes up about ~100 MB but it seemed like the server wasn't getting good info about my system and wasn't letting tasks in. I don't have 1000 GB available!

thanks a lot for your hint - it worked :-)))
I changed to "don't use more than 1000GB" - and after about 10 minutes a task came in.
This setting is really strange, isn't it? My system never could use more than 1000GB anyway, as the ramdisk is only 128GB.

So, as you say: the server obviously doesn't get correct info about the host.

KeithBriggs
Send message
Joined: 29 Aug 24
Posts: 10
Credit: 899,100,000
RAC: 13,469,212
Level
Glu
Scientific publications
wat
Message 61915 - Posted: 25 Oct 2024 | 0:29:41 UTC - in response to Message 61914.

Great news.

Post to thread

Message boards : Number crunching : no ATMMLs downloaded while many tasks from other project running

//