Starship

Starship is a tool to help simulate real chains, relayers and the interchain infra in a k8s cluster, locally, in the CI and on large k8s clusters.

Perfect for running internal devnets, and writing e2e tests. Starship helps take projects from development to production.

Directory Structure

• configs/ is a directory holding various kinds of configuration for running Starship
  • configs/local.yaml is configured specially to run locally
  • configs/devnet.yaml is a larger scale dev environment that will spin up nodes in k8s cluster and keep them running for a week (after a weeks time nodes might start to fail)
  • configs/ci.yaml light way setup supposed to be run in the CI runner itself
• scripts/: handy scripts to interact with Starship live here. These are standalone scripts that can be run anywhere
• tests/: Directory holding the e2e tests to run against the system
• Makefile: Handy commands go here

Getting Started

Setup dependencies

Checkout the docs or just run:

make setup-deps

Connect to kubernetes cluster

Connect to a k8s cluster. Follow one of the 2 steps based on your operating system

• 2.1.1: Spin up using kind cluster, for linux
• 2.1.2: Using docker-desktop, for mac

Run following to check the connection

kubectl get nodes

Startup

We use helm-charts for packaging all k8s based setup. This is controlled by the helm chart versions. Run the following to fetch the helm chart from the Makefile variable HELM_VERSION.

make setup-helm

Now you can spin up the local cluster with:

make install

Check the pods are in running state with:

kubectl get pods

Once all the pods are in Running state run port-forward command to get all ports forwarded locally:

make port-forward

Enjoy!!!!!

Teardown

Once you are done:

make stop

If you spun up the kubernetes cluster, then please stop it or if you used kind run make clean-kind

Run tests

Tests are designed such that one can re-run the same tests against an already running infra. This will save the cost of initialization of the infra.

Startup the cluster

make install

## check status of the pods
kubectl get pods

## Once the pods are up run:
make port-forward
 
## Run tests, can run this now multiple times as long as tests are running
make test

## Cleanup
make stop

Troubleshooting local setup

Currently, there seems to be some issues when running starship on a local system. This section will help clear out some of them

Not starting

If all or some of the pods are in Pending state, then it means that resources for docker containers are not enough. There can be 2 ways around this:

1. Increase the resources for your local kubernetes cluster.

• Docker Desktop: Go to Settings > Resources, increase CPU and memory

2. Reduce the resources for each of the nodes in configs/local.yaml file. You can look at configs/ci.yaml to understand the resource directive in the chains

• configs/ci.yaml uses very little resources, so should be able to run locally

NOTE: When resoureces are reduced or if the devnet has been running for a longer time, then the pods seem to die out or keep restarting. This is due to memory overflow. Will be fixed soon. For now one can just run

make delete

## wait for nodes to die out, check with
kubectl get pods

## restart
make install

Long startup time

Future Work

Dev-UX

• We will get rid of the whole scripts/ dir and replace it with a handy starship cli tool.
• Local infra spinup, specially on a Mac takes alot of time, this is something we will speedup much more
• We will have auto resource allocation, so manual troubleshooting is reduced
• Longer running pods or infra without issues.

Testing/Infra

• Add more tests based on requirements, port all existing adhoc tests to Starship
• Add build system into starship (premetive exists) to be able to build the current simapp from the branch
• Start from non-empty genesis
• Add the concept of jobs into Starship config, where we can run predefined jobs against the infra