[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Condor-users] Fresh install from a newbie

Hi all

I have a LOT of question, especially about the configuration I should use to run a condor grid on my network : first, I'll explain what I'd like to do, then I'll go through the "Global configuration file" and "Local configuration file" to ask which settings I should use (if you're not bored with my email until then...)

First, we use NFS on our network, so that each time a user goes on /nfs/XXXX, this refers to a NFS disk.
We also use NIS, so each user is referenced once on the NIS server, then each computer refers to this NIS server's /etc/passwd file.
I created a user "condor", who can access all the network, whose home directory is on the central manager, but accessible through the nfs with /nfs/condor.
The computers belong to a network called mylab.myfirm.com, like mycomp1.mylab.myfirm.com, mycomp2.mylab.myfirm.com, etc... but the correct email addresses are username@xxxxxxxxxx 

For now, we have around 20-30 machines to make calculations, and let's say 10 workstations (that we don't want to be "execute" machine : this might change later, but let's go step-by-step).

I plan to use a checkpoint server, which could be referred to as /nfs/checkpoint, so that it can be accessible through the network...

Now, I tried to install Condor 6.7.7, and started with the central_manager, following the manual : 
(from /home/condor on the manager)
>tar xfzv condor-6.7.7-linux-x86-glibc23-dynamic.tar.gz
>cd condor-6.7.7
>condor_ configure --install --type=submit,execute,manager --owner=condor 

OK, now everything seems fine : I tried some settings with the global config file, then tried to install an execute machine, but had especially problems with directories used (like RELEASE_DIR, LOCAL_DIR, etc...), that's why I'm now asking to those of you who understand my case...

Here are the questions : 
In the Global config file : 
- RELEASE_DIR : is by default set to /usr/local/condor : should I set it to the /home/condor (location of condor user's home) or /nfs/condor (the same, but accessible through NFS), or something else ?

- LOCAL_DIR : as is it linked to RELEASE_DIR, if it's on the NFS, that means I need a huge place (to store spool ??), right ?

- LOCAL_CONFG_FILE : seems to be automatically set during install... must it be accessible through NFS ? (in this case I have to change it...)

- USE_NFS : IMHO, should be True...

- LOCK : as my LOCAL_DIR (might) is on the NFS, I should set it as something like /tmp/condor, right ? or any better idea ?

- DEFAULT_DOMMAIN_NAME : should I set it to mylab.myfirm.com, or myfirm.com, or let it by default ?

Now, what about the local config file :
- RELEASE_DIR : does this have to refer to /nfs/condor (the home of the condor user on the manager), or must it be a local file, or anything else ?
- LOCAL_DIR : same question

- CONDOR_ADMIN : this was automatically set to root@xxxxxxxxxxxxxxxxxxxxxxxx, but it doesn't exist !! (or never checked...) : should I change it with the email address of condor (my email) ?

The final question is, when I setup other machines than the manager : do I have to create each time a /home/condor directory, and setup the condor client there ? do I have to configure also each time the Global Config file AND the Local Config file again ? Will some settings be automatically set by the installer ?

I really read through the manual, but as I'm quite new with cluster-thing, I might not have clearly understood some things, that's why my questions could seem so basic..

So, thank you if you take some time to help me.


Institut de Biologie Physico Chimique - UPR 9080
13 rue Pierre et Marie Curie
75005 PARIS

Tel : 01 58 41 51 70
Fax : 01 58 41 50 26