Add DigitalOcean provisioner

[klausenbusk]

This commit add a DigitalOcean provisioner and a flexvolume plugin
for DigitalOcean.

---

Finally, here we go :) The code has only been tested slightly on my k8s 1.8 cluster.

I'm not sure if I did the vendoring correctly (glide get github.com/digitalocean/godo + glide-vc --use-lock-file).. Edit: See commit message solved

/cc @andrewsykim

[k8s-ci-robot]

@klausenbusk: GitHub didn't allow me to request PR reviews from the following users: andrewsykim.

Note that only kubernetes-incubator members can review this PR, and authors cannot review their own PRs.

In response to this:

This commit add a DigitalOcean provisioner and a flexvolume plugin
for DigitalOcean.

Finally, here we go :) The code has only been tested slightly on my k8s 1.8 cluster.

I'm not sure if I did the vendoring correctly (glide get github.com/digitalocean/godo + glide-vc --use-lock-file)..

/cc @andrewsykim

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[andrewsykim]

cc @odacremolbap

[klausenbusk]

Ready for review, CI is green and the vendoring "issue" has been solved. The Go flexvolume plugin is maybe a big ugly, but it gets the job done..

[wongma7]

/lgtm

I will create quay.io/kubernetes_incubator/digitalocean-flexplugin and quay.io/kubernetes_incubator/digitalocean-provisioner and give you write permissions to them after we merge.

Thanks & sorry for the delay :)

[klausenbusk]

/lgtm

Great, I will rebase and remove the REVERT commit when I get home..

[klausenbusk]

Great, I will rebase and remove the REVERT commit when I get home..

Done and ready for merge :)

[klausenbusk]

@wongma7 are we waiting on anything? or maybe I'm just a bit impatient..

[klausenbusk]

 - [ ] Prevent k8s from scheduling more than 5 disks to a single droplet 

Probably solvable when Topology Aware Volume Scheduling get implemented.

[msau42]

I think for solving the max count scheduling issue, you're looking for a generalized version of kubernetes/kubernetes#53461

[klausenbusk]

I think for solving the max count scheduling issue, you're looking for a generalized version of kubernetes/kubernetes#53461

Interesting (but mostly related to internal volume plugin, it seems (?)), I will play around with whatever gets implemented first. MatchUnboundPVCs seems to provide enough functionality to implement à Volume Count Predicate like the internal volume plugin.

kubernetes/kubernetes#24317 also contain a few ideas.

[msau42]

The comments in the PR suggest an alternative using ConfigMaps to better support out of tree cloud providers.

[klausenbusk]

Thanks @wongma7 Do you have time to create quay.io/kubernetes_incubator/digitalocean-provisioner and quay.io/kubernetes_incubator/digitalocean-flexplugin ? :)

[wongma7]

@klausenbusk yes, I have sent you an invite also

Still trying to sort out and document how we should do versioning/release/auto-push for the repo, atm it needs a git tag to trigger build&push but that's not ideal. will create a github issue later.

[klausenbusk]

@klausenbusk yes, I have sent you an invite also

Great! I have just pushed both image as: quay.io/external_storage/digitalocean-provisioner:latest and quay.io/external_storage/digitalocean-provisioner:latest. The registry in the Makefile wasn't correct, so I have opened a PR to fix that: #520
BTW: It seems like I also forgot to add the provisioner to the Makefile, will do that later.

Still trying to sort out and document how we should do versioning/release/auto-push for the repo, atm it needs a git tag to trigger build&push but that's not ideal. will create a github issue later.

For now I'm just pushing as :latest, but it would be more ideal pushing with a version number, so backwards incompatible changes doesn't break a running cluster. Hmm...