exceeded disk limit?

0
0
Joined: 11 Feb 05
Posts: 11
Credit: 47160013
RAC: 0
Topic 196084

I noticed that recently most of my gamma pulsar tasks, and only these, are failing with:

02-Dec-2011 03:06:36 [Einstein@Home] Aborting task LATeah0069S_800.0_35450_0.0_0: exceeded disk limit: 20.99MB > 19.07MB

Like this one:
http://einsteinathome.org/task/260352041

Which is odd because I don't know what that limit is or why it is 21mb. Also I have 54GB free on C:, the only disk in the system, and boinc is set to use a maximum of 99gb, max of 99%, leaving 1 gb free.

This is the computer:

http://einsteinathome.org/host/3523267

Reading preferences override file
02-Dec-2011 00:46:30 [---] Preferences:
02-Dec-2011 00:46:30 [---] max disk usage: 55.34GB

This page: http://boinc.berkeley.edu/trac/wiki/Error/Aborting%20task%3A%20exceeded%20disk%20limit

...suggests this limit is a setting in the app/workunit itself?

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5872
Credit: 117845571652
RAC: 34733824

exceeded disk limit?

Quote:
.... boinc is set to use a maximum of 99gb, max of 99%, leaving 1 gb free.


There are three settings associated with disk usage.

Disk: use at most X GB
Disk: leave free at least Y GB
Disk: use at most Z % of total

Are you saying that your values are X = 99 and Y = 1? If so, that's fine and your problem may be what you have set for Z. If you have (say) Z = 50%, that means that BOINC will abort tasks if there is less than 50GB free (50% free) on a 100GB partition. As your normal activities outside BOINC start filling up your disk, BOINC will suddenly start showing what you are seeing if your TOTAL usage (including non-BOINC usage) starts exceeding the percentage you have set.

I've always found that values like X=10GB, Y=0.01GB and Z=99% prevent problems like those you are seeing. You might have to make X larger if you were supporting lots of projects and some had large storage requirements.

Cheers,
Gary.

Claggy
Claggy
Joined: 29 Dec 06
Posts: 560
Credit: 2699403
RAC: 0

RE: This page:

Quote:

This page: http://boinc.berkeley.edu/trac/wiki/Error/Aborting%20task%3A%20exceeded%20disk%20limit

...suggests this limit is a setting in the app/workunit itself?


I got myself some Gamma-ray pulsar search tasks to see what the task limits are, i make it 20Mb give or take:

Quote:

LATeah0072S_800.0_53550_0.0
hsgamma_FGRP1
23
60000000000000.000000
1200000000000000.000000
350000000.000000
20000000.000000

--inputfile LATeah0072S.dat --outputfile results.cand --f0start 736.0 --f0Band 64.0 --BskyPointFile skygrid_LATeah0072S_0800.0.dat --firstSkyPoint 53550 --numSkyPoints 50 --f1dot -1e-12 --f1dotBand 1e-12 --ephemdir JPLEPH.405 --Tcoh 524288.0 --toplist 10 --useWeights 1 --cohFollow 4 --inCohTimeRatio 35 --debug 1 --debugCommandLineMangling


LATeah0072S.dat
LATeah0072S.dat


skygrid_LATeah0072S_0800.0.dat
skygrid_LATeah0072S_0800.0.dat


JPLEPH.405
JPLEPH.405

The next question is, why is your host exceeding the task limits when your wingman isn't?

Claggy

0
0
Joined: 11 Feb 05
Posts: 11
Credit: 47160013
RAC: 0

RE: Are you saying that

Quote:

Are you saying that your values are X = 99 and Y = 1? If so, that's fine and your problem may be what you have set for Z.

I originally had them much more conservative, but I change them to see if it helped the problem; it didn't.

This is what it is set to now: Use at most 99gb, Leave at least 1gb, Use at most 90%.

But from Claggy's post it seems like this isn't the issue - there's a 20mb limit set in the task for some reason, and my machine is exceeding that. But why...

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5872
Credit: 117845571652
RAC: 34733824

RE: I originally had them

Quote:
I originally had them much more conservative, but I change them to see if it helped the problem; it didn't.


I saw the same type of problem some years ago and could suppress it with those three settings. It may well have been a function of an older version of BOINC which has been fixed in newer versions. It would appear to be unrelated to your current experience since you are using 90%.

Quote:
But from Claggy's post it seems like this isn't the issue - there's a 20mb limit set in the task for some reason ....


Presumably to stop a runaway process from creating a runaway output file and trashing your disk.

Quote:
... and my machine is exceeding that. But why...


Maybe you should try to confirm that. Many of your tasks that fail run for a couple of minutes before hitting the limit. 298 secs (5 mins) seems to be a popular figure. Yet, what is returned to the project is minuscule - just the error message. It might be possible to arrange for a new FGRP task to start and then jump into the relevant slot directory and see if you can see a file growing extremely rapidly. That might give some clue as to what is going on. I've just had a look in a slot dir for one of my tasks and nothing there is of any great size so a large problem file should be easy to spot.

Cheers,
Gary.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.