Here two handy scripts that give you an overview over the current jobs in the queueing system.
- jobstats gives a table with the number of jobs (running and pending) for all users.
- myjobs lists the number of your jobs in the system.
Month: July 2010
SSH host key changed
As the operating system was replaced completely, the SSH host key of majorana changed as well.
If you get an error message from your ssh client upon trying to connect to majorana, follow the instructions to delete the cached key (usually in ~/.ssh/known_hosts).
Scheduler Problems
Over the last two days, the main scheduling daemon of SGE had major problems, essentially eating up all memory and then silently crashing afterwards. I think I fixed it; you should be able to submit jobs without problems now.
Back online
The new cluster is online. Please note that I’m still testing it, so expect some bugs. Currently only the second rack is online (c1-x and c2-x).
The following things change for you, the users:
- After logging is as usual (ssh majorana), you will end up on the new submit node “submit-1-0″. It’s a dedicated host for compiling, data analysis and job submission. That way, the main server does not get bogged down by user programs. On this submit node, you can work as usual.
- All software packages were updated, you need to recompile all
your programs. It might well be that some packages are missing,
so tell me if you need something. - The main servers (majorana, and the two storage servers oss0 and
oss1) are off limits for normal users. - We have 8 new nodes!
Ah, and don’t pay any attention to the cluster status page right now; it
does not work atm.