[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Condor-users] no scheduling with Condor-G 6.7.19 and Globus 4.0.2
The answer to my questions will be evident for the one who knows but... let's quickly describe the situation:
* I have one running Globus grid with the GT4.0.2; At this point, I still have to specify the target of a job when submitting it with globusrun-ws.
* So I decided to use a scheduler over my grid. I choosed Condor because I already have one efficient "pure" (without globus) running condor (6.6.11) pool. So I installed a new Condor of the last version available.
I was able to submit some jobs without condor-G but I felt on the followign problems:
1) I am still forced to explicitely set the target in "grid-ressource" (but I would let condor choose the best one). So there is for me no other real benefit than the queue management.
2) I am unable to submit a job on grid nodes where several job-managers of the same type are registered. That is really annoying because my sample grid has 3 nodes, with one DefaultIndex on each, and two indexes upstream their registrations to the third.
This let me hope I "just" forgot one part of the grid, I mean, a way to advertise to condor what is available on the grid. Hawkeye seems to be a solution since it has a grid-monitoring module. I tried to do hand-made advertising but it was result-less.
Thus, my questions are:
* Do my expectations match any existing solution? (I hope so, but...) and is hawkeye a good candidate?
* Do I have to install hawkeye on everey node or can I just monitor the node that gathers the others indexes?
I thank any experience feedback, any advice, any useful link.
Thank you very much!
iBELGIQUE, exprimez-vous !