I see no reason to re-image the node. Generally you can choose the base folder (ie. /usr/local/) of your blat installation and synchronize each node with the head node with the old but power "rsync" tool. It works like a charm in my Apple cluster. On 2/3/05 5:47 AM, "Juan Carlos Perin" <bic at genome.chop.edu> wrote: > I'm installing and testing out MPI and BLAS on my local G5 desktop, in order > to attempt to run LINPACK. Ultimately, I would like to run Linpack on our > Apple cluster to get some figures... > > Installing BLAS is taking a long time. And I assume the libraries are > required on every node for linpack to run properly. Is there an easy way to > do this? > > Also, this plays into a few other issues. Installing MPI blast seems to > have been relatively straightforward, although the same issue was asked, and > we decided that copying the final binaries to each node would be sufficient. > This seemed to work, but still leaves some doubt in my mind. Should we be > compiling on each node? I know BLAS is a set of libraries, so this is not > as simple as copying a few binaries, but there must be an easier way... I'm > naïve to this area. > > This also begs the question, would it be best to configure out head-node > exactly how we want it with all the correct libraries, and then simply > re-image each node with a boot disk image of the headnode? > > > Thanks, > Juan Perin > > > _______________________________________________ > Bioclusters maillist - Bioclusters at bioinformatics.org > https://bioinformatics.org/mailman/listinfo/bioclusters