QA » openQA Project

Sometimes, a test failure occurs because there's no network connection in the SUT. Then we hit our post-fail hook, which probably tries to upload logs for diagnosing the problem...over the network...

I'm doing a sorta hacky workaround for this in the Fedora tests: have the post-fail hook check if it can ping the host IP, and if it can't, try and dump the contents of text logs via the serial console. That seems at least better than nothing.

It occurred to me that it may be interesting to do a more generalized implementation of this in os-autoinst - perhaps as a new command similar to upload_logs, or implement it within upload_logs based on whether it can ping the host IP, something like that? Does anyone think this sounds like a useful idea?