commit | 88de9f82b1c99543bd7cedb123bf8868758b60f3 | [log] [tgz] |
---|---|---|
author | Bogdan Caprita <caprita@google.com> | Wed Jun 01 11:23:22 2016 -0700 |
committer | Bogdan Caprita <caprita@google.com> | Wed Jun 01 20:59:05 2016 +0000 |
tree | d0df413875de89f5766f581c5195b0958159f3a4 | |
parent | 69a358670e180d51a9eefffc5752295ed06db233 [diff] |
allocator: use the instance kubernetes name as the canonical instance key The goal is to break the implicit relation between the kube name and the mount name of the instance. The mount name is generated for the instance at creation time and then stored with the instance. Deriving one from the other after that point is fragile, since the format of either could change in the future (or we could have zero or more mount names per instance). Most of the changes are mechanical and result in simplifications. In particular, the handlers are easier for users to work with than full mount names. There are two parts that become slighly more complex: 1. The dashboard handler still needs the mount name; instead of extracting it from the deployment json (which would incur an extra kubectl command), we pack it in the request url together with the instance handle. To prevent undue access, we sign the pair. 2. The allocator cmd-line tool "list" command now has two modes, showing just the handles or full details about each instance. While at it, also remove the transitional logic that filled in the mount name and blessing names for instances that didn't have this persisted in their deployment annotations. Change-Id: I4380ad4d3376fa64961048838815fdc2f8e71a3a
This repository contains a reference implementation of the Vanadium APIs.
Unlike the APIs in https://github.com/vanadium/go.v23, which promises to provide backward compatibility this repository makes no such promises.