Talking about modules and build systems, take a look at Easybuild: http://hpcugent.github.io/easybuild/
This is not a direct answer to the original poster's question - which has been admirably dealt with in previous answers. On 26 May 2015 at 23:22, Michael Gutteridge <[email protected]> wrote: > I'll plus-one your two-cents there. > > We're using an NFS mounted application file system where we have requested > tools- it's organized by tool name and version. Modules has been an > absolute revelation for keeping things sensible when maintaining multiple > versions of software. > > We do use Puppet for managing the nodes themselves, including the versions > of system software that are installed (/usr and below). Fair warning: > Puppet is fairly dependent on packaging for deploying software. There are > add-on modules that allow installation from tarballs... those do a pretty > good job, but I haven't used them extensively. > > What you may look into at the same time is build management and automation > for your unbundled tools. We don't have anything at the moment and I'm > *really* not looking forward to the next OS upgrade. > > Best > > Michael > > > On Tue, May 26, 2015 at 2:56 PM, Chris Dagdigian <[email protected]> wrote: > >> >> Environment Modules (http://modules.sourceforge.net/) or the more modern >> derivatives are pretty darn universal in cluster environments where a >> shared filesystem can be used to host an application/tools tree. Do one >> install and one module file and it's instantly available across the >> cluster. Does a great job of allowing you to host and offer many different >> versions/permutations of the same application. >> >> After that you get less universal and odder more DIY setups. Some groups >> will roll software into custom .rpm, .pkg or .deb packages depending on >> their OS and will do all of their custom/internal software as binary >> packages hosted on a central app repository server. Lots of hacks involving >> passwordless SSH scripts run on remote nodes to replicate files or >> dump/unpack a tarball representing a tool. >> >> And finally the cool DevOps kids are using Puppet, Chef, SaltStack, >> Ansible, Cfengine, etc. configuration management / orchestration toolkits >> -- those are often overkill for simple software installs but they *excel* >> at maintaining and managing configuration states across many different >> systems. If your entire cluster is Puppetized one could easily come up with >> a puppet-based method for installing or managing the installation of >> software >> >> My $.02 of course >> >> >> >> >> Trevor Gale wrote: >> >>> Hello all, >>> >>> I was wondering what solutions other use for easy software installation >>> across clusters. Is there any method that is generally accepted to be the >>> most effective for making sure that all the software is consistent across >>> each node? >>> >>> Thanks, >>> Trevor >>> ___________ >>> >> >> _______________________________________________ >> Beowulf mailing list, [email protected] sponsored by Penguin Computing >> To change your subscription (digest mode or unsubscribe) visit >> http://www.beowulf.org/mailman/listinfo/beowulf >> > > > _______________________________________________ > Beowulf mailing list, [email protected] sponsored by Penguin Computing > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > >
_______________________________________________ Beowulf mailing list, [email protected] sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
