Closed rhettg closed 14 years ago
After much work and twisted pain, ssh connection failures are handled fairly gracefully.
There isn't much in the way of tests for this, mostly it was manual work with a vm host just to see what twisted did. Might not be reasonable to add much unit testing for this, I"ll have to think about it more.
Tron handles running a job by SSHing into a machine and waiting for it to complete.
I have no idea what it will do if the machine disconnected.
At a minimum, we should mark the job run as some 'unknown' state and optionally leave the job in a state where it can try again once the machine comes back.