[Swift-devel] nearly, but not quite optimal performance on ranger from ranger under coasters

[Michael Wilde]

On 4/29/09 7:55 PM, Mihael Hategan wrote:
> On Wed, 2009-04-29 at 19:40 -0500, Glen Hocky wrote:
>> Tried to rerun (identical everything) but am now getting the errors in 
>> this logfile
>>
>>> /home/hockyg/public_html/swift_logs/ranger_from_ranger/oops-20090429-1920-q6r3oph1.log
>> Errors are:
> 
> org.globus.gram.GramException: The job manager detected an invalid
> script response

I vaguely recall this error being causes when the job time exceeded the 
limits of the queue time, or some similar error.

Look to see if you got PBS errors in your home dir, or clues in the gram 
logs on the target site.

One case was that the calculated time needed for the coaster jobs 
exceeded the queue limit.

Of course, it could be something entirely different.
> 
> I'm getting the same when I do a simple job submission. Apparently gram
> on ranger is broken again.

Did you do this from communicado?

I was getting gram err 74 there, but gram 12's everywhere else. When I 
added a osg client into my .soft the errors went away.

If this was communicado, there is something else going on there I think. 
It seems that the globus 4 package may have been damaged in some way.

> I've said this before, and I'll say it again:
> 
> Unless TG gets its act straightened out about the reliability of gram
> installations (and it hasn't for about 5 years now), running on a single
> site is going to be a problem. The only reasonable solution I can think
> of is running on multiple sites, and letting the swift fault-management
> do its job.

I agree that this is what we want to have working well as the default 
way of using Swift. That's what we were trying a bit ago with ranger, 
abe, and queenbee, and the intent to expand the list.

We've got to get some expiring & ending accounts renewed, which is going 
to slow us down a bit, but we'll get back to testing in that mode.
> _______________________________________________
> Swift-devel mailing list
> Swift-devel at ci.uchicago.edu
> http://mail.ci.uchicago.edu/mailman/listinfo/swift-devel