Grendel - grendel.chm.bris.ac.uk
NB grendel is currently being reconfigured- information on these pages may be out of date and features described may not currently be implemented!
For an account on this cluster email chem-unix@bristol.ac.uk with your UoB username, name of supervisor and preferred login shell (if you have a preference).
For support email chem-unix@bristol.ac.uk.
Passwords- See this page for information about usernames and passwords on grendel.
Queuing system- Grendel uses a queuing system for running jobs. See this page for important information about how it works.
Each year you will have to renew your account on grendel for it to
remain open. You will receive an email about this in January each year,
please reply to this email stating whether or not you wish your account
to remain open.
If/when you leave Bristol, please contact the system administrator and
provide a new email address so that you can continue on the grendel
email list.
Notice under the Data Protection Act
File storage space is provided on grendel solely for your research data and results. Private files should not be kept on grendel as they are not held confidentially. Access to your files will be provided to members of staff, and to others if the appropriate member of staff requests it.
The System
Grendel is a beowulf cluster, that is, a cluster of computers
connected together. The master computer is called the front end,
the others are called nodes.
The front end is connected to the departmental network, the nodes are
not. When you login to grendel, you login to the front end, which is
used for compiling, editing files etc, not for running jobs.
Run jobs by submitting them to the queuing system, which will
distribute them to the nodes.
The operating system is Linux.
The users' filestore is backed up nightly with a combination of full
and incremental backups. The full backup happens every 2-3 weeks at the
weekend and takes two days, so backup frequency is lower on those weekends.
Nodes:
- 12 nodes, g-0-0 to g-0-11, with 2 Xeon 1.8 GHz processors on switch1.
Each node has 1.0 GB RAM and 32 GB of scratch space on /tmp. - 16 nodes, g-1-0 to g-1-15, with 2 Xeon 2.8 GHz processors on switch2.
Each node has 2.0 GB RAM and 32 GB of scratch space on /tmp. - 10 nodes, g-2-0 to g-2-9, with 2 Xeon 2.8 GHz quad-core processors
(ie 8 cores per node) on switch3.
Each node has 8.0 GB RAM and 400 GB of scratch space on /tmp.
Note that the RAM and scratch space is for the node as a whole, not
for each processor.
Nodes b01-b12 are connected to the front end (and
each other) by one ethernet switch (switch1), nodes b13-b29 are
connected by a second ethernet switch (switch2) and nodes b30-b39 are
connected by a third ethernet switch (switch3), and so there is no
direct connection between the three groups of nodes.
The nodes on switch1 all have 100Mb/s connections to the switch and the
switch has a 1Gb/s connection to the front end.
All connections on switch2 and switch3 are 1Gb/s.
The Front End
Only secure shell connections are allowed
to this computer.
Samba connections to grendel can only be made from inside the
department.
The front end is for editing files, compiling etc. Run jobs
by submitting them to the queuing system. Short tests
(1 minute maximum) may be carried out on the front end. Longer tests and any
jobs must be submitted to the queuing system. Tests and large compile jobs
on the front end MUST be run at low priority using nice -n 19 <command>
i.e. to run a program called a.out, use the command
nice -n 19 a.out
Processes running which fail to meet these requirements will be killed.
There are two main shells available, bash (an extension to the Bourne shell) and tcsh (an extension to the C shell). If you wish your login shell changed, contact the system administrator (the user command for this, ypchsh, has a bug).
- If you use tcsh, the system reads the files /etc/csh.cshrc
and then /etc/csh.login and then the files in your home directory
.tcshrc (or .cshrc if it is not there) and .login. Edit the files in
your home directory to customise your sessions. You can read the system
files to see what options have already been set. When writing shell
scripts you should note that the -f option does not read .tcshrc, but
will read .cshrc if it is there.
- If you use bash the system reads the files /etc/bashrc and /etc/profile and then the files in your home directory .bash_profile and .bash_login if they exist, or .profile if they do not. Edit the files in your home directory to customise your sessions. You can read the system files to see what options have already been set.
When you login you will be shown how much space has been used and how much is available. Read the file Diskuse in /home, /home2, etc to see how much space people are using, and if you see that you are using a large amount, do something about it before the situation gets out of hand.
To ease transfer of files between grendel, dirac and bohr:
dirac's /home, /home2 and /home3 partitions are cross-mounted on grendel
as /dirac/dirac, /dirac/dirac2 and /dirac/dirac3.
bohr's /users and /users1 partitions are cross-mounted on grendel
as /bohr/bohr and /bohr/bohr1.
These are automounts, that is, they will automatically mount when
you try to access them, and unmount when you have finished.
From the front end you can, if necessary, login to the nodes using the commands ssh g-1-13 (for example).
Micellaneous
- Some useful utility packages installed on the front end.
- The queuing system for running jobs on grendel.
- Programs available on grendel.