Admin Tips

From wiki
Jump to: navigation, search

Simple commands on all nodes, giving node name

  • Gives the running kernel version of all the nodes:
for i in node{1..10}; do ssh $i 'echo -n "$HOSTNAME: "; uname -a';done

New users

After having set up, here is some text that can be sent to a new user:

This is a unix/linux based system of 11 high speed machines for processing bioinformatics datasets. The frontend or gateway machine is called marvin, while the other 10 are called nodes, numbered 1 to 10. Most operations are done using command line, although it's possible to use Graphical Interface, not very many programs have this available.

Further details here: 

How to use: 

Other useful links:

Three useful links about command-line bioinformatics:

Compiling software

Red Hat Developer Toolset

Seeing as the cluster runs with RHEL 6 which, though updated continuously, was released in 2011, there are various things it's incapable of. Primary among these is a gcc compiler with the C++ standard C++11, which alot of software is now using. THis requires gcc version 4.8 instead of the 4.4 RHEL6 uses. The solution is the poorly named developer toolset, which Red Hat makes available. To invoke it:

scl enable devtoolset-2 "bash"

This will open a new bash subprocess where the gcc will be version 4.8

There is also GCC 6.3 installed too.

source /opt/rh/devtoolset-6/enable

Note: loads it's own python and perl.


  • Very unconventional manner of installation, creates symlinks from the buildtree.
  • this means that testing must be done via a users who had no hand in compilation.
  • definitely needs RH Developer Toolset

Procedure: clone from git-tree, run autoconf, and the configure step should be something like this:

./configure --prefix=/usr/local/Modules/modulefiles/tools/harvest-tools/gitv0_ebf7ecc --with-protobuf=/usr/local/Modules/modulefiles/tools/protobuf/gitv0_0622030 --with-capnp=/usr/local/Modules/modulefiles/tools/capnproto/0.5.3