[GE users] looking for feedback: new web interface to qstat based on XML to HTML transform

Chris Dagdigian dag at sonsorol.org
Tue Apr 26 19:15:14 BST 2005


In order to keep the sge users mailing list on-topic I've created a 
mailing list for people who want to talk details about xml-qstat and/or 
get access to the raw sources for the cgi's before I get the time to 
formally package something up. If this topic interests you feel free to 
sign up at:

http://bioteam.net/mailman/listinfo/xml-qstat-d

People who don't want to join a mailing list but have a questions should 
feel free to contact me directly so we can keep signal-to-noise ratios 
high for the pure SGE mailing lists.

Since the list did not exist until a few moments ago I'll answer the 
questions as best as I can here:

Currently the stylesheet will 'transform' any XML that gets sent to it. 
Massive output (including the multi-megabyte examples that people sent 
me for testing) will all get rendered into a massive web page. I'm 
hampered by not having access to a massive cluster right now for testing.

There are 2 ways to adjust this:

  o Server side: The CGI program that calls qstat can be hard coded with 
a different set of qstat arguments instead of the default "-f -xml 
-explain aAcE" syntax. We can maybe even improve the CGI such that the 
specific qstat call can be user-adjusted on the fly to select queue 
subsets or "show me only running jobs" etc.

  o XSL stylesheet: XSL is pretty darn powerful and I've only just 
scratched the surface of what it can do. It is possible to code logic 
into the stylesheet such that it would do things like "transform only 
the first N nodes of the XML input". XSL even supports basic named 
functions and even parameter passing so you could presumably pass a 
parameter on the fly to the XSLT engine that results in limiting or 
constraining the HTML output to a certain size regardless of what XML is 
fed into the engine.

Regarding execd spool outout:

  o It does require access to the execd/messages file. If this is not 
possible because spool logs are non-local than you can trivially edit 
the XSL stylesheet to comment out the section that inserts the execd 
links. Rapidly changing the appearance of the HTML output is exactly why 
I find this whole grid engine XML -> HTML transformation stuff exciting.

-chris



Sean Dilda wrote:
> This looks nice.  I actually had a user forward me your post to 
> bioclusters list as he thought it was a really good idea.
> 
> I wonder if you've given any thought to how this will work on larger 
> clusters.  My cluster has over 300 hosts in SGE and over 500 queues 
> (going bye the <queuename>@<hostname> convention).  As such, it'd mean 
> users would have to scroll through 500 lines of queue info before they 
> got to the job info part.
> 
> Also, where does it get the execd information from?  Does it require the 
> execds have their spool directories on NFS?  If so, is there any easy 
> way to disable the execd functionality for those of us who use local 
> spooling for execds?
> 

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list