[Condor-users] Condor, Condor-G and Hawkeye

[Jean-François SMIGIELSKI <jf-smig@xxxxxxxxxxxxx>]

> Hello Jean-Francois,
> 
> I'm sorry I can't help but I was wondering if you could tell me which parts
> of Condor software you installed on the globus grid nodes? Do you have a
> full Condor installation on each and every grid node? Because just the
> condor_advertise executable doesn't seem to suffice for advertising.
> 
> I'm having the same problem with manual advertising, in addition to having
> to install Condor on each grid node. I don't know about Hawkeye but I will
> go out and look for some info :-)
> 
> Dragan Sunjka

Hello.
Please excuse the latency.

The grid nodes only have the GT4.0.2 installed, and they are host-monitored with ganglia. Only one of these nodes has Condor-6.7.19 installed with master, collector, negociator and shedd running.
I found it pretty easy to parse the ganglia generated XML output to build a condor ad, and to automatize it with cron. But I miss some parameters in my ads.

I am still Iooking for a better tool, but I wouldn't have to install condor on all Grid machines (I shouldn't have to) or hawkeye (I still don't know what it i used for) because it seems so big for my little task, and it would cause more and more maintenance.

So, any explanation about what does hawkeye (and how it does) would be helpful.

Thanks a lot,

JF Smigielski


PS: I wrote this mail without responding to the original, and then I broke the response suite. Here's the original:

> Hello, list!
> 
> I have some small questions about hawkeye.
> 
> My context:
> - a globus grid (GT 4.0.2) of linux machines
> - a condor pool (6.7.19, not yet 6.7.20:-) managing one usual machine (Linux) and the previous
> Globus Grid.
> - I "manually" send ads for the grid nodes to the condor_collector, and these machines appear
> when I call condor_status. I am able to automatically dispatch jobs using condor-G. But I still have
> do advertise by hand.
> 
> I would have some automatic state collection for my grid. Hawkeye seems to be a solution, but I
> don't find a good online documentation about it.
> 
> My questions:
> * Is hawkeye the tool I need to collect my grid's nodes states?
> * If it is true, do I have to install hawkeye on all the nodes? Can I instead install it on the main
> index node?
> * Do you have any useful links, or any good advice?
> * Why does hawkeye_collector listen to the same default port than condor_collector? This cause
> condor_master to be unable to spawn condor_collector or a bind error.
> * So, Does hawkeye's daemons replace some Condor daemons?
> 
> I understand the "which does what" of the condor daemons, but I have a lack with Hawkeye. It
> seems that Hawkeye and Condor have overlapping concerns.
> 
> Thank you (the list) for your help.
> Thank you (the condor-team) for the great job you've done.
> 
> Jean-François Smigielski.


________________________________________________________________________
iBELGIQUE, exprimez-vous !
http://web.ibelgique.com/