I'm trying to executing high computation tasks on top of dispy nodes. Actually my cvm has 8 cores.
When I submit the jobs I set the zombie interval as 5 mins. But even the computation takes more than 5 mins, the processes were not marked as zombie process, so that I couldn't be able to use that node for subsequent processes.
Any idea to terminate those long running processes and enable that node for future jobs?
When I submit the jobs I set the zombie interval as 5 mins. But even the computation takes more than 5 mins, the processes were not marked as zombie process,
A node / scheduler / client is a zombie if it is not reachable after a period (deemed dead). This is based on pulse messages that are exchanged (at 'pulse_interval'). This has no relation to computation times. You may wanto use 'terminate' method of jobs in the client program to kill processes that are taking longer than certain time.
Ignoring invalid request from '<head node ip address : port>'
It is not possible to figure out why you get this message; may be you can investigate what is running at given address:port?
And also why dispy scheduler is sending empty byte to each worker nodes? is it ping messages?
Scheduler (nor any other component in dispy) send empty byte messages. Ping / pulse messages (like most, if not all, messages) are dictioaries.