Meeting Notes for 16 April 2008
Present: Bob, Tom L., Tom J., Mihael, Liz, Joao, Eric;
Absent: Dan K, Mike W., Randy, Kris
Release of Refactored Code
Problem with load-balancing script
The first part of the telecon was spent fixing a problem with the script to do load-balancing being deleted; this code resided in Tibi's area; Mihael fixed the problem. He also had to restart the Tomcat servers on www14 and www15.
Problem with SSH LIbrary
We still have the problem this library causes with concurrency of jobs submitted to the i2u2 cluster using SSH.
Liz asked if there was some way to limit the number of jobs that go to i2u2 cluster so we could still have it, but only so many at a time? Would this avoid the concurrency issue?
Mihael: It might reduce the incidence of the problem.
MIhael wants to keep execution on the i2u2 cluster. You can see that it hangs and you can start another one. The hang does not gobble up resources.
Mihael: There might even be some educational value for kids and teachers to see that things don’t always work; this is the environment researchers have as well.
Mihael: we want to get feedback from users; – We want them to do some problem solving. They need to be strategic.
Deployment
We decided to deploy on the weekend. Bob will send out an email and include the bug report link.
Other Issues
Job lifetimes
The time an analysis stays in the list is two days. We discussed increasing the lifetime of jobs to four days so that students could submit jobs on a Friday and still see it on Monday; We realized that this would interfere with work on the weekend when we want to do releases. Students lose jobs when we redeploy.
Mihael said that it is hard to track the jobs at a sysadmin level - The api was not open enough to do this.
Other suggestions:
- Have a pop-up, with a warning – if this job is alive for two days.
- calculate an estimate the time of run and disallow it.
- put something on the Execution Choices page.
Bob will test these:
- Geometry upload confirmation
- Upload progress wasn’t working
VDS Execution
Mihael wants to keep VDS in the list of execution choices.
Report on ATLAS e-Lab from Joao
Project was set back because CERN asked him to do things. He’s back to Amelia. Tom and Dan will come in May. He wants to have it ready by then. He wants it hooked to SWIFT. Change in philosophy. Problems with the Firefox engine under Mac and Linux. He is in the middle of redesign. Now using QT,("cute"), a powerful GUI library with a lot of networking modules, database modules. Webkit is the engine used in Safari.
Java wrapper for the 3-D engine. A regular java interface. The CERN exhibit service wants it. In early May – new user interface – all the features except the interface to SWIFT. This work rearranges the items in the Amelia part of the diagram. See MeetingNotes20080305#Diagram_V2.
Tom and Dan – May 12-14th. Liz asked if Mihael could come then? He will be in Taiwan May 3-9th so he probably won't be able to take the time away from his other work, but he can have a conference call with them.
LIGO
Eric and Mihael need to work on things together. (e.g., saving plots)
CMS
Mihael – has to work on this, but he does not know when he can work on this. Currently the website does not have the sections on the various studies.
-- Main.LizQuigg - 23 Apr 2008