Giridhar Pemmasani

Once you get source from git, you can either copy files to where your current installation is, or you can generate sdist file with `python setup.y sdist` and then install...

Can you compress each log file and send each as separate email to me directly? The problem seems to have happened before first 'Ignoring invalid reply'. The size of about...

You can save log files by redirecting (may want to use 'tee' as well to know the problem happened to stop); e.g., 'dispynode -d | tee /tmp/dispynode.log'. Please send logs...

I haven't used it that way, but I think when you use with tee, it runs as daemon, so no input is possible. However, it should serve fine. At the...

Let me test and get back to you.

Just did a test and it works fine. Note that with 'tee', log is buffered so you may not see output until more output is produced. Also, you can give...

As discussed above, how about not using SharedJobCluster (and dispyscheduler)?

If there is a considerable delay when submitting jobs, it is possible node may have closed computation (see 'zombie_interval' option to dispynode). Update committed above may help.

There is no reason to use threads with dispy. There is no benefit with using more than thread (due to Global Interpretter Lock with Python) and likely it is more...

`NodeAllocate` instances can be used as `nodes` to customize allocation of resources on each node, e.g., as: cluster = dispy.JobCluster(compute, nodes=[dispy.NodeAllocate(host='*', cpus=1)]) It can also be subclassed for more control:...