[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Condor-users] alternative 6.7.18 RPM for linux-x86
- Date: Tue, 28 Mar 2006 12:06:46 -0600
- From: Erik Paulson <epaulson@xxxxxxxxxxx>
- Subject: [Condor-users] alternative 6.7.18 RPM for linux-x86
I've been maintaining an alternate RPM for the Condor releases for some
internal projects for a while, and I've decided I should release them. The
binaries used to create the RPM are identical to the official release,
the only thing that's different is the packaging and the configuration. We
use these RPMs in an environment with no shared filesystem, and a custom
Linux install managed with cfengine, so the RPM assumes everything is
local to the machine. It is meant to be the RPM for a worker node.
It should run on nearly any RPM-based x86 linux distribution, but
more recent releases (Fedora Core 5, for example) will need the
compat-libstdc++ RPMs installed as well.
The big differences in the configuration between "official" Condor and
* Condor is installed directly into /usr, not /opt/condor-release
* There is no condor user, instead the daemons use 'daemon' as their EUID
* Configuarion is stored in /etc/condor/condor_config. Do
not edit this file, make all changes to
condor_config.local. condor_config.local will not be overwritten in an
upgrade, but /etc/condor/condor_config will be completely replaced and
any changes to it will be completely lost. In condor_config.local, you
should only have to set three variables: condor_host, collector_name,
and hostallow_write. It is empty after it is installed, you must
edit this file.
* You can finally actually use rpm --upgrade
* condor_configure and condor_init are not used at all and are removed
* The VMx_USER setting is used. The RPM creates execute users
* The number of CPUs detected is capped at 4, to prevent running
out of VMx_USERs. If there are installations with more than 4 CPUs
I will up this in future releases
* EXECUTE_LOGIN_IS_DEDICATED is turned on to prevent lurker processes.
* The startup script in /etc/rc.d/init.d is completely rewritten, and much
* Local files like log files and the execute directories for jobs
are in /var/condor. You may want to make that a seperate partition.
I have signed the RPMs. If you come to Condor Week (which you should, since
it'll be a good program) I will give you a copy of my public key fingerprint
The RPMs are here:
You should consider them 'experimental'. If you're using them and having
problems with condor, try the official releases before asking
condor-admin for help.