I removed the disks that weren't in use at all. But those were only 4 1GB ones. However, there are 3 100GB ones in use, and I have no idea what app/deployment using them (except that it's Kubernetes).
Even more interesting, and relevant to the issue I ran into: https://github.com/drone/drone-runtime/issues/44
I just found this forums post, which explains how Kubernetes support works, where to find what, how to debug it, and how to contribute: https://discourse.drone.io/t/contributing-to-drone-for-kubernetes/3159
Just FYI: We have access to Digital Ocean's Kubernetes beta. Load balancers over there support Let's Encrypt out of the box and cost $10/month. 2GB nodes also cost $10/month. There's no additional charge for managing either of those with Kubernetes (same as GKE).
Just FYI: I think we may want to run this in the more traditional way on andromeda, until this stuff is stable enough. At least if we see that it breaks regularly on GKE.
What's wrong with simply deploying them to a PV and mounting that as /custom?
OK, good. So then what's missing here is to actually copy the entire custom folder there, right?
create ConfigMaps for the entire folder instead of just options/label