Re: [Condor-users] DAGs and Error msg "Too open files"

["Michael O'Donnell" <odonnellm@xxxxxxxx>]

Kent, thanks. I have changed my DAG and submit files for the DAG by not 
specifying log files and I will be testing this today. I suspect this will 
help a lot but it seems like this would be a problem if I had a much 
larger pool. Our pool is not that large, but I try to always keep the 
scalability in mind as I learn and implement Condor for my employer.

thank you for the help,
mike







From:
"R. Kent Wenger" <wenger@xxxxxxxxxxx>
To:
Condor-Users Mail List <condor-users@xxxxxxxxxxx>
Date:
12/09/2011 08:30 AM
Subject:
Re: [Condor-users] DAGs and Error msg "Too open files"
Sent by:
condor-users-bounces@xxxxxxxxxxx



On Fri, 9 Dec 2011, Michael O'Donnell wrote:

> I am running a DAG in our pool (all machines including the central 
manager
> run on a Windows OS (XP, 7, R2) with about 200 available machines for 
this
> specific job and my DAG reports an error which changes its status from
> Running to Idle (that is to say the jobs are running fine but the
> condor_dag.exe produces an error).
>
> The error log file the DAG produces this:
> 12/09/11 06:22:48 Can't open "ExtSimVal_DAG.dag.dagman.out"
> dprintf() had a fatal error in pid 9684
> Can't open "ExtSimVal_DAG.dag.dagman.out"
> errno: 24 (Too many open files)

I'm not an expert on the Windows side of things, but one thing you *can* 
do is reduce the number of file descriptors DAGMan uses by having all of 
your node jobs use the same log file.  The easiest way to do this 
(assuming you're running a fairly recent DAGMan -- 7.5 or later should be 
good, I think) is to just not specify a log file in your submit files, and 

let DAGMan generate a default log for you.

(One note: we are working on changes to DAGMan that will limit the number 
of file descriptors it uses at any given time, no matter how many log 
files are used by the DAG node jobs.  But those changes are not ready 
yet...)

Kent Wenger
Condor Team
_______________________________________________
Condor-users mailing list
To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/condor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/condor-users/