search

diux-dev / cluster

train on AWS
75 stars 15 forks source link

Trouble running ray_integration/launch_simple.py #1

Closed robertnishihara closed 7 years ago

robertnishihara commented 7 years ago

A couple comments. I had to add

export AWS_DEFAULT_REGION=us-west-2

Then had to run python launch_simple.py. Running ./launch_simple.py didn't quite work because launch_simple.py is not executable.

Then launch_simple.py seems to do an infinite loop for me. Output so far is attached. The only error I see is 

mkdir: /temp/tasklogs/simple/0: Permission denied

output.txt

yaroslavvb commented 7 years ago

Can you see if this commit works? It should now write to /tmp instead of /temp, I'm assuming /tmp is writeable which is true on my MacOS and Ubuntu 16.04

yaroslavvb commented 7 years ago

ps, the reason I chose /temp instead of /tmp is because system reboot wipes /tmp, but it turned out tasklogs are not that useful so it should be ok to lose them sometimes

robertnishihara commented 7 years ago

Yep, it works now. I was able to ssh just fine. Nothing much is supposed to be installed on the machine, right?

yaroslavvb commented 7 years ago

Nope, it installs everything from scratch, it's the "INSTALL_SCRIPT" parameter inside "launch_simple.py" and "launch_ray.py"