Table of Contents
<box green| IMPORTANT> Please always check with Yaxun before you follow any instructions on this page as 'root' or 'super user'</box>
Creating new accounts
- Login as root on any of the nodes, say, sep101.
- Edit /etc/group, add the user to appropriate group
- Edit /etc/passwd, add the user information, make sure the UID hasn't been used by any other user
- Edit /etc/shadow, if the user want to use the same password as on 'koko': copy the corresponding shadow information from /etc/shadow on koko, make sure you copied the correct line; another choice would be issue 'passwd user' to create a new passwd
- scp /etc/group /etc/passwd /etc/shadow to all nodes
- make home directory for the new user on all nodes
- change the ownership of the directory to the new user
- make scratch directory on all nodes and change their ownerships
- You are all set
ssh/scp/sftp without password
- Generating keys. On the local machine (koko or pompei, for example) other than the cluster nodes, issue “ssh-keygen -t rsa” to generate RSA keys (or DSA keys if you want). Your RSA public key will be generated in ~/.ssh/id_rsa.pub
- Creating .ssh directories. Create directory ~/.ssh on the remote machines if it does not exist. Then issue “chmod 700 .ssh” to change the permission of this directory.
- Copying keys. Copy your public keys from the local machine (~/.ssh/id_rsa.pub) to all the remote machines (~/.ssh/), and rename it to “authorized_keys”, then issue “chmod 600 authorized_keys” to change the permission of this file.
- You should be all set.
rsh/rlogin/rcp without password
- Login as root on any of the cluster nodes, check /etc/xinetd.d/rsh and /etc/xinetd.d/rlogin, make sure 'disable = no'
- Restart the 'xinetd.d' service by 'service xinetd restart'
- Make sure /etc/securetty contains rsh and rlogin, if you want root to rsh without password
- Create .rhosts in users' home directories and add the IP addresses of the hosts that are allowed for rsh/rlogin connection
- Alias rsh to /usr/bin/rsh. By default, rsh is pointed to /usr/kerberos/bin/rsh, which uses kerberos 5 and it is not working properly for rsh, so you always get annoying message when connecting with rsh. Aliasing to /usr/bin/rsh solves this problem, since it tries kerberos 4 first. For csh/tcsh user, add alias rsh '/usr/bin/rsh' to your '.cshrc'; For bash user, add alias rsh='/usr/bin/rsh' to your '.bashrc'
- Try 'rsh host' to see if it works
- scp the user's '.rhosts', '.cshrc' or '.bashrc' to his home directory on other nodes
Note: rsh/rlogin has security problems, so ssh without password is always preferred.
Lists of software on sep100s and sep400s
- Intel compiler suite (10.1.015)
icc : /opt/intel/cc
ifort: /opt/intel/ifc mkl : /opt/intel/mkl
- SEPlib
binaries: /opt/SEP/SEP6.5.4 sources : /opt/SEP/seplib-6.5.4
- Seismic Unix
/opt/SU
- FFTW
fftw2.1.5 library: /opt/FFTW/lib/
fftw3.1.2 library: /opt/FFTW/lib/
- MPI
/opt/MPI/mpich-1.2.7p1
Lists of healthy nodes
- sep100s
sep101 sep102 sep103 sep105 sep106 sep107 sep108 sep109 sep110 sep111 sep112 sep113 sep114 sep115 sep116 sep117 sep118 sep119 sep120 sep121 sep122 sep123 sep124 sep125 sep126 sep127 sep128 sep129 sep130 sep131 sep132 sep133 sep134
- sep400s
sep402 sep406 sep407 sep408 sep409 sep411 sep412 sep413 sep415 sep416 sep418 sep420 sep421 sep423 sep425 sep426 sep429 sep430 sep431 sep432