[GE users] Manually merging two accounting files?

jcolvin jcolvin at sfwmd.gov
Thu Feb 5 21:03:10 GMT 2009

Never mind, that's the correct new behavior as of 6.1. 

-----Original Message-----
From: Colvin, Joshua 
Sent: Thursday, February 05, 2009 3:44 PM
To: 'Alex Chekholko'; users
Subject: RE: [GE users] Manually merging two accounting files?

Thanks Alex. Done and it seems to work fine, although I did notice
"qstat -f" now only shows me my running jobs, not every running job.
If I'm another user "qstat -u $USERNAME -f" will show that user's
jobs okay. "qstat -f" used to show me all jobs by default. I'm not 
sure if this is 6.2 behavior or a side-effect of merging the
accounting files.

-----Original Message-----
From: Alex Chekholko [mailto:chekh at pcbi.upenn.edu] 
Sent: Thursday, February 05, 2009 2:43 PM
To: users
Cc: Colvin, Joshua
Subject: Re: [GE users] Manually merging two accounting files?

On Thu, 5 Feb 2009 14:31:17 -0500
jcolvin <jcolvin at sfwmd.gov> wrote:

> Hello,
> We had one 6.0u12 SGE installation, recently replaced (not upgraded)
> another running 6.2u1. 
> We still have the old data available from backups. I am wondering if I
> can merge the accounting files 
> from both? It looks like overlap from identical job IDs would be the
> only issue. If I sed the new accounting
> file's job IDs to numbers greater than the old accounting file's job
> IDs, and manually create a larger, unused 
> number in the new SGE's spool/qmaster/jobseqnum file, would this work?
> I can launch multiple "qacct -f" queries to gather data from both and
> merge them, but I'm wondering
> if there's an easier way, by just manually merging the two backend
> files?

I recently upgraded from 6.1u3 to 6.2u1 via the cluster clone method
described in that excellent video.

I opted to restart the jobids at 1 with the new SGE version.

AFAIK, the clone upgrade simply copied the accounting file and it is
still being appended to.

So if I do 'qacct -j JOBID', I get back two results; the old job and
the newer job.  This is not a problem for us as we typically search by
date and/or user, and don't care about jobs from long ago.

I think you can easily update the seq_no of your new install to be
larger than the highest job number from your old install, if that
solution is palatable for you.



To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

More information about the gridengine-users mailing list