Workers in distributed mode fail to run `virtualenv`

[ludovicchabant]

In Distribute.py (line 496), FunkLoad tries to create a new virtual environment by running python virtualenv.py /path/to/temp. At least in my case, it failed to run it because virtualenv.py couldn't be found. I'm not sure what caused this, my workers were all fresh EC2 instances running the standard Ubuntu AMI, and the only thing I did on them was to install pip and virtualenv.

I worked around the issue by modifying the code locally to just run virtualenv /path/to/temp directly.

I'm not sure why FunkLoad goes to all the trouble of finding the virtualenv script and all that, instead of just running it anyway?

[bdelbosc]

Hi,
you don't need to install virtualenv on the slave, only install virtualenv and paramiko on the master as described in the doc:

sudo aptitude install python-paramiko, python-virtualenv

I will commit some fix today regarding the --distribute-workers opts that fails when submitting pwd or explicit ssh port.

[ludovicchabant]

Mmmh I had paramiko and virtualenv installed on the master. So unless installing virtualenv on the slave somehow disrupts funkload, then something was still not working right... I'll try again soon.

[jerri]

I am trying this also and the problem persists even with funkload-1.17.0. The created virtualenv.py on the worker-servers seem to have problems to find pip and die. Therefore the needed virtual environment is never set up correctly.