Google Cloud Platform is a cluster, in the <a href="http://www.urbandictionary.com/define.php?term=Cluster&defid=1424405" rel="nofollow">http://www.urbandictionary.com/define.php?term=Cluster&defid...</a> sense.<p>Amazon eats their own dogfood with AWS services, while Google does not, and no amount of marketing is going to make up for that. Every time I give it a try, I find random bugs everywhere, and I know there are not enough internal engineers feeling the pain to get those bugs fixed.
We found that we could still go into their console, SSH into the boxes (from the browser-based thing), and reboot the boxes from there. When they came back up, they worked. May have just been good luck, just happening to come back up on hardware that's not behind the bad routers or whatever it is.<p>Edit: Also, we found that our instances were reachable (they mostly provide a JSON-based API over HTTP), in the sense that they were getting the incoming HTTP/HTTPS traffic that they normally do. But the responses were not getting back out to whoever requested them... lost on the way back out or something.
All our status VMs are unreachable: <a href="https://cloudharmony.com/status-for-google" rel="nofollow">https://cloudharmony.com/status-for-google</a>
On any post incident reports - shall we ever come out of using the most common and ambiguous technology lingo "network issue". If it were identified a network issue (Route black hole, prefix hijacking, resources depletion or consistency issues etc) then you probably know enough to elaborate on the specifics.
Preliminary cause is described here: <a href="https://status.cloud.google.com/incident/compute/15045" rel="nofollow">https://status.cloud.google.com/incident/compute/15045</a>