Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 324654

Summary: PBS launch configuration should make better use of queue information
Product: [Tools] PTP Reporter: Greg Watson <g.watson>
Component: RM.PBSAssignee: Project Inbox <ptp-inbox>
Status: RESOLVED FIXED QA Contact:
Severity: enhancement    
Priority: P3 CC: arossi, roland
Version: 4.0.2   
Target Milestone: ---   
Hardware: Macintosh   
OS: Mac OS X   
Whiteboard:

Description Greg Watson CLA 2010-09-07 09:31:19 EDT
The PBS RM launch configuration just seems to provide an attribute field labeled "destination" for the destination queue. There doesn't appear to be any way for a user to find out the acceptable values for this attribute. However the PTP model provides a "queue" model element that is designed to be used to populate a dropdown combo (or similar) so that the user has some idea of the queues that are available, and this should be being populated by the RM (if it is working correctly). I suggest looking at providing such combo for the PBS launch configuration.
Comment 1 Albert L. Rossi CLA 2010-09-07 11:15:50 EDT
Sure, this shouldn't be too difficult to do.  I'll take a look.
Comment 2 Albert L. Rossi CLA 2010-09-10 10:57:54 EDT
The Launch Tab for PBS now displays the list of machine queues available as a combo box, values derived from the IPQueue elements sent during RM model definition.

The auxiliary method for populating the box is improperly encapsulated (it belongs in the Wizard Page), but this would have required an API change.

Also, "destination" must be a fixed item in the PBS client's attribute name vocabulary for this to work.

Committed to HEAD and 4_0.
Comment 3 Greg Watson CLA 2010-09-10 14:36:47 EDT
I'd suggest a couple of other things. This field should probably be first in the list of attributes, since it is probably the most important. How are the attributes ordered? Second, "destination" seems a strange designation. Is this PBS standard? I would have thought "Queue" would have been better, or at a minimum "Destination".
Comment 4 Albert L. Rossi CLA 2010-09-10 16:15:04 EDT
(In reply to comment #3)
> I'd suggest a couple of other things. This field should probably be first in
> the list of attributes, since it is probably the most important. How are the
> attributes ordered? Second, "destination" seems a strange designation. Is this
> PBS standard? I would have thought "Queue" would have been better, or at a
> minimum "Destination".

The user can order them any way she wants, through the template.  The default templates certainly can put it first.

All the attribute names are exactly the same as what PBS specifies in its internal API.  The shell command is -q.

The description field reads: 

"Designation of the queue to which to submit the job." 

Hover over destination, and you will also see the message:

Format: queue[@server]

Since we are processing attributes dynamically now, we can't really do arbitrary mappings because that would require knowledge of the attribute names ahead of time; however, since the client does need to know about this particular attribute, the mapping could be hidden from the user and "queue" used instead in this case.
Comment 5 Albert L. Rossi CLA 2010-09-10 16:25:16 EDT
> Since we are processing attributes dynamically now, we can't really do
> arbitrary mappings because that would require knowledge of the attribute names
> ahead of time; however, since the client does need to know about this
> particular attribute, the mapping could be hidden from the user and "queue"
> used instead in this case.

On further reflection, PBS followed the original model of using the pbs API names, but since the client actually contructs the shell scriipt, it really isn't necessary.  The XML file on the proxy side could arbitrarily change these names, so long as the mapping to the attribute flags is known somewhere.

And, this reminds me that this latter mapping was not addressed in our original plan, so I need to bring it up with Roland and Benjamin.