sosGetCommandOutput doesn't actually use timeout about sos HOT 5 CLOSED

I think this gist would work perfectly in solving our timeout issues with long running processes or those that become defunct in sosreport's presence

https://gist.github.com/1306188

from sos.

If we're going to solve this we should solve it generally (fs access via addCopySpec* is vulnerable to the same problems). I'm interested to know if anyone has ideas here but the last time I looked into fork/exec/wait in python it seemed difficult to write this kind of thing robustly. I have one solution using an external C binary that's simple and covers the command cases well but if there are other suggestions we should evaluate those too.

from sos.

It's funny you guys are talking about this... The other day I started playing around with trying to figure out how to make this work and came up with this:

https://gist.github.com/7b75afd647a8dac49ef1

There are a lot of problems with trying to make one implementation. My implementation probably doesn't work on windows, but it will let you execute via a shell (which is what most calls expect).

A c-extension won't work for jython, but we might be able to use it for platforms that do and come up with a jython implementation to fall back on.

from sos.

I'd be interested to know what happens when the child gets into trouble - is a single fork and pipe enough to protect us from hanging inside p.communicate()?

from sos.

If you mean to use /usr/bin/timeout as a prepended shell argument on systems where it exists then we can do that under jython as well.

You can try my gist out like so

python timeout.py 10 "cat /dev/random > 10_seconds_of_dev_random"

Though, I'm not sure that is a good test either, as I doubt that will fill up any pipe buffers. p.communicate() is supposed to buffer into memory so we could always get into trouble there if we are drinking from a firehose.

from sos.