Hi Massimo et al,
I will try to reply here in one shot to most of your emails.
Massimo Sgaravatto wrote:
> What is the scenario you are interested in ?
> Torque batch master installed on the LCG-CE node ?
Our scenario is trying to cover quite a few requirements together,
supplying in one setup 4 different targets: lrms, lcg-CE, creamCE, arc01
Separating the lrms from lcg-CE is a natural choice in this case,
partly for scalability and partly for maintainability.
Jeff Templon wrote:
> I have set up a CREAM service parallel to our two lcg-CEs, for which there
> is a fourth machine serving as the LRMS server. I am now making this setup
> "sustainable" as there were indeed some hacks necessary to get it to work...
> I will make the documentation on this setup public when I am done.
Hi Jeff,
your feedback would be greatly appreciated, by many others as well I guess!
Stephen Burke wrote:
> Both of those services are being phased out, which is why they haven't
> been ported to SL5. [...] Similarly the lcg-ce is being replaced by CREAM -
> at the moment it seems that CREAM isn't quite ready to take over completely
[...]
Burke,
I think we are missing here a roadmap that basically accepts that lcg-CE
and cream-CE are going to be supported together for 10 months or so.
I believe it is unrealistic to claim the switchover will happen overnight
or that lcg-CE@SL4, as it has, can cope with our needs.
We already had this discussion in the past about how long it takes to upgrade
the whole infrastructure, as of now, ~3 months would be the absolute minimum.
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind05&L=LCG-ROLLOUT&D=0&P=1501207
markus wrote:
> we have been considering porting the LCG-CE to SL5 for a while,
> did most of the work, but decided not to release it.
[...]
> Except that we are really encourage sites to run a CREAM-CE in
> parallel so that we can switch soon....
Markus,
What was the reason for not releasing? this is important information!
Having it out and calling it a beta gives important advantages in that
we can give it a try and pool our (site) experience on it.
Douglas McNab wrote:
> Torque client and maui client versions between the SL4 CE's (torque-2.3.0)
> and SL5 Torque server (torque-2.3.6) are different
Yes, thanks for the maui tip and all the other nice sysadmin-relevant details,
I can already expect it can prove troublesome, maui key is in our checklist.
We'll also try to reach the latest possible bug-patched versions in our setup,
so it was good to see that others have worked in the same direction.
> Any tweaks we have made are managed by cfengine
This is extremely interesting, since we use cfengine, too. (but w. SL5!)
Can you supply some details by direct email? grid _at_ cscs dot ch
Daniela Bauer wrote:
> The machine was setup that way after I had a run in with our system
> manager who refused to put anything of the SL4 variety on our new
I think your system manager is doing you a favor ;-)
I come to realise that you have deployed sge rather than torque/maui,
so your fan-base in this list is reduced compared to what I assumed.
Still interesting for some, but not option for us due to another dependency.
thank you all for your time reading this,
Fotis
|