Daniel Fagnan
Daniel Fagnan
A separate task should handle pinging each peer in the cluster and getting an `ok` response back. This should be configurable, too. This could be implemented as a method on...
Trying to figure out a proper ownership scheme for the tcp streams. Now that they are all copyable, we can have multiple copies in different tasks, which is pretty nice....
Need to split up the simulation functionality into it's own package in the same repo. This should include the required functionality to spin up simulations against an executable, with it...
Setup vagrant to run the simulation tests locally. All the simulation functionality only runs on linux.
Document the protocol is great depth. I've split the docs in multiple files under the `docs` folder. This will allow people unfamiliar with the concepts to grasp it fairly quickly....
There should be heavy simulation testing to simulate 100s and 1000s of nodes within a cluster. This should ensure the system's stability and correctness. There should be a centralized proxy...
The CI infrastructure should include a full set of [Jepsen](https://github.com/aphyr/jepsen) tests. This should ensure that the properties and design decisions are fully correct and that the trade-offs explained are accurate.
Recently hit a bug where if the child process doesn't exit cleanly retool will leave it orphaned leaving you to force close the child process yourself. This was hit with...
Could we simply return an `error` type instead of panicing? This seems like a recoverable error to me. cc @achille-roussel ``` panic: (*Message).Finish or (*Message).Requeue has already been called goroutine...