container.execute() not working

I have flask application (flask + socketio + gunicorn), and pylxd not working correctly on it. Problem in container.execute() method.

For example

print(0)
container.execute(['ls'])
print(1)
container.execute(['cd', '/var/www/html'])
print(2)
container.execute(['service', 'nginx', 'restart'])
print(3)

Output: 0

Please help me 😦

pylxd = 2.2.10 lxd = 3.18

[ajkavanagh]

Thanks for the update. The warnings are because LXD has grown some new attributes on some of the models, and pylxd is unaware of them. You can silence them by setting PYLXD_WARNINGS=‘none’ in the environment before running the app which loads pylxd.

This was added in #361 – by default one warning is printed for each attribute known about; there is an ‘always’ setting which prints the warning for every subsequent model load.

I think you are probably hitting #379 which I’m looking at. An (ugly) work-around is to use subprocess.Popen and just run lxc exec directly through that; but this only works if it is on the same machine and the user has the same permissions as the pylxd user that is being used.

I may provide that as an alternative exec method, but it has the above significant issues.

[ajkavanagh]

Popen(...) by default connects the stdin, stdout and stderr to the parent process’ (i.e. your script’s) stdin / stdout. Therefore, the print("OKKKKKK") has no where to go. So if the command is still running on the container, then it is still holding the stdout and thus the print in the script will be buffered. What you really want to do is let the script run in the container that starts the service as a daemon and then exits. At least, I suspect that is what is going on.

You only need to use Popen(..) if you want to communicate with the process as it is running. i.e. send it input, etc. Then you use the .communicate() method after setting up the stdin, stdout, etc. to PIPEs to do that communication. If you are running one-shot commands, it’s easier to use subprocess.check_call() and friends.