New GFRegression URL is unresponsive and diffbrowsers doesn't use a request timeout.

[graphicore]

Well, there are two issues A) http://35.238.63.0 (use here) is not responding to GET (any) requests. I'm not sure where to report this.

B) and this is more important: diffbrowsers should use a timeout when using requests.get or requests.post, which is not the default.

e.g. in GFRegression from diffbrowsers.gfregression

    def _validate_instance(self, url):
        """Confirm instance_url is a working instance of GFRegression"""
        request = requests.get(url, timeout=5)
        ...

Otherwise, in case of A) the workers in the dashboard hang just forever waiting for an answer.

The timeout can be bigger than 5 seconds of course, it's more important that it fails eventually than quickly.

[graphicore]

BTW. I'm having the same issue with $ gftools qa from the command line.

[m4rc1e]

Thanks for reporting this. Usage has increased by ~500% in the past two months.

GCE recommends I upgrade the instance.

I'll do this now.

[m4rc1e]

Ok it's back. I'm going to check the logs to see what's happening.