discovery: bad discovery endpoint Unable to GET token

[mikesimos]

I 'm getting discovery: bad discovery endpoint errors on two existing discovery endpoints.
A get request to the discovery url yields: 400 (Bad Request) Unable to GET token
However, I was able to create new discovery endpoints that work as expected.
Any known issues?

[colonha]

I'm reporting here same problem. Cannot access discovery.io and new machines are not able to join the cluster.

[Flygsand]

Same issue here right now. Did anything change recently?

[colonha]

Looks like CoreOS (Redhat) updated the API to a newest version removing some features (Like DELETE) and remove some (Not sure if all) exiting keys.

Since I was using these removed features I decided to build a private etcd discovery service by myself.

[lucab]

/cc @philips I think this may be related to the current ongoing work for https://github.com/etcd-io/discovery.etcd.io.

[philips]

The discovery service did change. See https://groups.google.com/forum/#!topic/etcd-dev/q-pN4PiGY_o

Unfortunately we had to delete all existing tokens in the migration. However, discovery tokens should only be used for initial boot strapping and were periodically deleted in the past. So, I am surprised to hear there are issues.

The new discovery.etcd.io code base is in a PR here etcd-io/etcd#10627 and if you are having issues please file a detailed issue on https://github.com/etcd-io/discovery.etcd.io.

@colonha In particular I would be interested to hear what API calls you were using exactly. My intention was not to break the existing API but I ported it to a new backend so it is possible something not covered by a test broke.