Message boards : ATLAS application : Atlas task failing Computational Error with exceeding disk limit
Message board moderation

To post messages, you must log in.

AuthorMessage
Cphipps

Send message
Joined: 1 Aug 14
Posts: 15
Credit: 6,817,923
RAC: 5,629
Message 34907 - Posted: 8 Apr 2018, 0:34:39 UTC

Hello All. Several of the LHC (Atlas) with the following msg. How can I increase my disk limits. My disk preferences are set for unlimited.

4/7/2018 3:58:15 PM | LHC@home | Aborting task AELKDmo60OsnlyackoJh5iwnABFKDmABFKDmBRdTDmABFKDm0OQ0Wn_0: exceeded disk limit: 8296.34MB > 7629.39MB
ID: 34907 · Report as offensive     Reply Quote
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 674
Credit: 43,168,217
RAC: 16,180
Message 34908 - Posted: 8 Apr 2018, 9:18:12 UTC - in response to Message 34907.  

I think that this error comes from the internally set limit in the task. This is set by the server when task is created. So you cannot change it.
ID: 34908 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 15 Jun 08
Posts: 2386
Credit: 223,032,772
RAC: 137,063
Message 34909 - Posted: 8 Apr 2018, 10:14:17 UTC - in response to Message 34907.  

At least a couple of your tasks ran as 8-core VMs.
This is most likely too high for your hosts.
Beside that, an 8-core setting is rather inefficient.

You may try to limit the VMs to a much lower core setting (1-4) via the project's web preferences or a local app_config.xml.
Be aware that changes to the web preferences only affect WUs that are downloaded after the change.
ID: 34909 · Report as offensive     Reply Quote
Erich56

Send message
Joined: 18 Dec 15
Posts: 1686
Credit: 100,473,584
RAC: 104,204
Message 34911 - Posted: 8 Apr 2018, 12:02:02 UTC - in response to Message 34909.  

Beside that, an 8-core setting is rather inefficient. ... You may try to limit the VMs to a much lower core setting (1-4)
BTW, I am wondering which core-setting is indeed the most efficient one.
I guess I remember some postings where it was said that 4-core would be most efficient. Is this still valid?
ID: 34911 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 15 Jun 08
Posts: 2386
Credit: 223,032,772
RAC: 137,063
Message 34912 - Posted: 8 Apr 2018, 12:23:36 UTC - in response to Message 34911.  

Beside that, an 8-core setting is rather inefficient. ... You may try to limit the VMs to a much lower core setting (1-4)
BTW, I am wondering which core-setting is indeed the most efficient one.
I guess I remember some postings where it was said that 4-core would be most efficient. Is this still valid?

In principle: less cores are more efficient, as I tried to explain in a simplified generic example.
Running more cores per task makes sense if you don't have enough RAM to satisfy all cores you want to use or if you want to have lower task runtimes.
ID: 34912 · Report as offensive     Reply Quote
PHILIPPE

Send message
Joined: 24 Jul 16
Posts: 88
Credit: 239,917
RAC: 0
Message 34916 - Posted: 8 Apr 2018, 15:57:55 UTC - in response to Message 34912.  
Last modified: 8 Apr 2018, 16:12:38 UTC

In addition to this (what is the best core configuration?) , the answer is not absolute but depends on your computer ( OS age and type , kind of cpu , amount of ram available , ssd or hdd ,...) and also on your way of crunching ( host dedicated or not , full or partial time to crunch ,only one project at a time or several ones , with or without one or two core idles to give easier the maintenance by your OS, the isp quality and your bandwith, the skillness to understand where a bottleneck appears and how to solve it,...).

I found this plot on the dashboard which displays the average cpu efficiency of all the kind of cores configurations executed by all volunteers (for only jobs with success) :



over a sliding period of one month.

As we can't select options to make a deeper inspection and separate parameters to study their influence on the results , you have to consider it as a merge of global work with Boinc.There is no possibility to give accurate results for a particular host type.

The only thing you can do is to compare your situation with the average provided in this plot and appreciate your position (above or below the average).

In order to estimate the data volumes which are used to calculate the above average efficiency:



over a sliding period of one month.

Independently of the graph , i notice that linux native application has a better efficiency than virtual box for a same number of cores and enable 12 cores simultaneously with good results when virtual box is limited to 8 cores because of its lower efficiency.
(but it requires more skillness and knowledge to be installed).
ID: 34916 · Report as offensive     Reply Quote

Message boards : ATLAS application : Atlas task failing Computational Error with exceeding disk limit


©2024 CERN