UofL CS Condor Pool
From Ben's Writing
Contents |
Introduction
Installation
Linux
Missing Libraries
On our Linux machines, compat-libstdc++-33 is required for Condor to work. To install compat-libstdc++-33 run the following command:
$ sudo yum -y install compat-libstdc++-33
Hostname Problem
Seems the `hostname` binary does not work as is expected by the Condor configuration scripts. As a result, the /etc/hosts file requires a quick update:
# Do not remove the following line, or various programs # that require network functionality will fail. 127.0.0.1 localhost.localdomain localhost ::1 localhost6.localdomain6 localhost6 142.66.140.42 ichor.cs.uleth.ca ichor
That is, we added the fully qualified hostname to the file.
Directories
Currenty there is no scratch area on the CS machines, which means we are forced to use the /var partition for our lock, spool and execute directories. The log directory will be stored in ~condor/hosts/$(HOSTNAME)/log, one per machine.
This is a problem because a malicious (or ignorant) job could fill the partition and bring the machine down.
I've recomended that newly installed machines have a scratch partition of aproximatly 10GB be allocated for this purpose. This should solve the problem in the future. This is tentatively scheduled for the summer, as the lab machines cannot be changed mid-semester.
Solaris
Hostname Problem
Like Linux, Solaris has some problems with hostname resolution on our network. Because we are running Solaris 10, this problem can be easily fixed by adding two new files: /etc/nodename and /etc/defaultdomain.
- /etc/nodename: should contain the fully qualified domain name. For instance, artsy, would written as artsy.cs.uleth.ca.
- /etc/defaultdomain: should contain the domain name: i.e, cs.uleth.ca.
We do all of this because the solution used on the Linux machines cannot be implemented as simply on Solaris. For details click here.
Statistics
To check the usage statistics of the pool, click here. (Note: you must be on site to visit this page.)
Environment
Bellow is the base environment required to use Condor on our pool. Note that the MPICH2 and PVM binaries are optional.
## set the root directory for all of our binaries BINARIES_ROOT=/local/raid1/condor/binaries BINARIES_SYSTEM=`uname`-`uname -p` ## add the Condor binaries to the path CONDOR_BINARIES=$BINARIES_ROOT/$BINARIES_SYSTEM export PATH="$CONDOR_BINARIES/bin:$CONDOR_BINARIES/sbin:$PATH" ## add MPICH2 to the path MPICH2_BINARIES=$BINARIES_ROOT/mpich2-$BINARIES_SYSTEM export PATH="$MPICH2_BINARIES/bin:$MPICH2_BINARIES/sbin:$PATH"
