[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] checkpoint on java



On Jul 31, 2006, at 10:35 AM, a..darrigol wrote:

I intend to do a pool of windows and unix machines to run java
programs.
Is it possible to checkpoint a job on the java universe? If
not, is it possible to use the standard universe with a java
program by using condor_compile?

My problem is that the windows machines will be switch off
every night so if java jobs can't checkpoint they will start
again everyday from the begining and they will never finish
the execution...

In theory, you could run java jobs in the standard universe by condor- compiling the jvm. The jvm would have to obey all of the restrictions of the standard universe, most notably no threads. I don't know if anyone's successfully tried it.

I have another problem concerning condor_vacate and condor_off.
After condor_off i have this message:
Sent "Kill-All-Daemons" command to master birch
but in fact the command kill no daemons.
It s the same with condor_vacate that never stop the jobs on
the machine. Can we use condor_vacate with a job that can't be
chepointed like a vanilla job?

The commands were probably refused because the sender wasn't authorized to issue them. Unfortunately, these errors aren't sent back to the condor_off/vacate command from the daemon. If you look in the daemon's log, you should see a line saying that the command was denied.

+--------------------------------+-----------------------------------+
|           Jaime Frey           | I used to be a heavy gambler.     |
|       jfrey@xxxxxxxxxxx        | But now I just make mental bets.  |
| http://www.cs.wisc.edu/~jfrey/ | That's how I lost my mind.        |
+--------------------------------+-----------------------------------+