General Information

This section contains general information about the Atmospheric Sciences UQAM computing cluster and diagnostics environment.

The UQAM/SCA Linux cluster called marvin is generally used to run the GEMCLIM model.

Students of Colin Jones and Laxmi Sushama are expected to store their modelling results on one of the three UQAM/SCA storage servers called st1, st2 and st3. These machines can also be used for post processing and debugging. Please contact Katja Winger (Katja.Winger@ec.gc.ca) if you need storage space on these servers.

IP Addresses and Hosts

The Linux cluster marvin can be accessed from anywhere. Whereas only one (st1) of the three storage servers (st1,st2,st3) can be accessed from outside. The other two (st2,st3) can only be accessed from marvin or st1.


Machine Description IP Address Address Type Host Name
Cluster 132.208.132.89 external marvin
Diagnostics/Storage 132.208.132.18 external st1

internal
(from marvin or st1)
st1
st2
st3

Data Storage on marvin and st1, st2, st3

All of the disks on marvin and st1, st2 and st3 are RAID-enabled. This reduces the chances of data loss in the event of a disk crash; however, it also makes disk space more valuable. 

The first level storage on the cluster marvin is a set of FibreChannel disks mounted at /local/fiber1. This 1.7T RAID has the fastest I/O access on marvin, and should be used only at runtime by the model. No final storage of data should be done on /local/fiber1!!! The second level storage on marvin are two larger (8T and 5T) RAID units made of slower SATA disks mounted at /local/sata1 and /local/sata2. This is where the model output will appear after all post processing and diagnostics, and it is from here that data should be transferred as soon as possible to longer-term storage. For students of Colin Jones and Laxmi Sushama this would be one of the servers st1, st2 or st3. This means that all other students will have to use their own servers *** as there are absolutely no long term disk allocations on marvin ***

Machine Mount Point Size (G) Purpose
marvin /local/fiber1 1700 Runtime Storage (Level 1)
/local/sata1 8300 Temporary Output Storage (Level 2)
/local/sata2 5500 Temporary Output Storage (Level 2)


Changing Passwords on marvin

The username and password information on marvin (the cluster) uses a Network Information System to distribute information across nodes. This means that running passwd to change your password will not produce the desired result. To change your password run the command:

    yppasswd

Customizing the Environment

Your environment can be customized to your preferences (for example, choice of compilers and sets of aliases) using the ~/.profile_usr file that resides in your home directory. The standard Unix method for customizing the environment for Borne shells and derivatives is to modify the ~/.profile file; however, the RPN Library uses this file, and in turn "sources" the ~/.profile_usr file. Modify this file exactly as you would the ~/.profile file, and you can customize your environment as you like.


Interactive jobs on marvin

Interactive jobs should only be submitted from node cn149, never from the headnode!
You can log on to the node with:

    ssh cn149


Batch jobs on marvin

Batch jobs must only be submitted using 'soumet'.
And please always use the 'soumet' from the environment and never use 'qsub' directly.


Manuel

Also have a look at the French manual for marvin.




Author: Ron McTaggart-Cowan
Last update: August 2008 (Katja Winger)