The following dependencies are required for local development:
- Go >= 1.17.5
- Tilt >= 0.20.8
- Any of the local Kubernetes clusters supported by Tilt.
We strongly recommend minikube >=
v1.21.0 with the kvm2 driver.
- Tilt will use Minikube's embedded Docker server. If Minikube is not used, a local instance of Docker / moby-engine >= 19.03 is required.
See the Tilt docs docs on how to set up your local cluster - it won't take more than a few minutes to set up! Example minikube invocation, adjust limits as needed:
minikube start --cpus=8 --memory=8G --disk-size=50G --driver=kvm2
npm wants to set up an insane number of inotify watches in the web container which may exceed kernel limits. The minikube default is too low, adjust it like this:
minikube ssh 'echo fs.inotify.max_user_watches=524288 | sudo tee -a /etc/sysctl.conf && sudo sysctl -p'
This should work on Linux, MacOS and Windows.
By default, the devnet is deployed to the wormhole
namespace rather than default
. This makes it easy to clean up the
entire deployment by simply removing the namespace, which isn't possible with default
. Change your default namespace
to avoid having to specify -n wormhole
for all commands:
kubectl config set-context --current --namespace=wormhole
After installing all dependencies, just run tilt up
.
Whenever you modify a file, the devnet is automatically rebuilt and a rolling update is done.
Launch the devnet:
tilt up
By default this runs a network consisting of one guardian, two ganache (Eth) chains, a Solana test validator, an Algorand sandbox, and LocalTerra for both Terra Classic and Terra 2. If you want to work on non-consensus parts of the code, running with a single guardian is easiest since you won't have to wait for k8s to restart all pods. See the usage guide below for arguments to customize the tilt network.
Watch pod status in your cluster:
kubectl get pod -A -w
Get logs for single guardian node:
kubectl logs guardian-0
Restart a specific pod:
kubectl delete pod guardian-0
Adjust number of nodes in running cluster: (this is only useful if you want to test scenarios where the number
of nodes diverges from the guardian set - otherwise, tilt down --delete-namespaces
and restart the cluster)
tilt args -- --num=2
Run without all optional networks:
tilt up -- --algorand=false --evm2=false --solana=false --terra_classic=false --terra2=false
Tear down cluster:
tilt down --delete-namespaces
Once you're done, press Ctrl-C. Run tilt down
to tear down the devnet.
This tutorial assumes a clean Debian >=10 VM. We recommend at least 16 vCPU, 64G of RAM and 500G of disk. Rust eats CPU for breakfast, so the more CPUs, the nicer your Solana compilation experience will be.
Install Git first:
sudo apt-get install -y git
First, create an SSH key on the VM:
ssh-keygen -t ed25519
cat .ssh/id_ed25519.pub
You can then add your public key on GitHub and clone the repository:
git clone git@github.com:certusone/wormhole.git
Configure your Git identity:
git config --global user.name "Your Name"
git config --global user.email "yourname@company.com"
Your email address should be linked to your personal or company GitHub account.
After cloning the repo, run the setup script. It expects to run as a regular user account with sudo permissions. It installs Go, Minikube, Tilt and any other dependencies required for Wormhole development:
cd wormhole
scripts/dev-setup.sh
You then need to close and re-open your session to apply the new environment.
If you use ControlMaster SSH sessions, make sure to kill the session before reconnecting (ssh -O exit hostname
).
Start a minikube session with recommended parameters:
start-recommended-minikube
You can then run tilt normally (see above).
The easiest way to get access to the Tilt UI is to simply run Tilt on a public port, and use a firewall
of your choice to control access. For GCP, we ship a script that automatically runs tilt up
on the right IP:
scripts/tilt-gcp-up.sh
If something breaks, just run minikube delete
and start from scratch by running start-recommended-minikube
.
VSCode's SSH remote development plugin is known to work well with the workflow described above.
IntelliJ's remote development backend is reported to work as well. Just install Jetbrains Gateway on your local machine, connect it to your remote VM via SSH, and pick the latest IntelliJ release. Your local license, keymap and theme - if any - will be used automatically.
Projector should also work for clients that can't run the native UI locally (if you want to code on your VR headset, smart toaster or Chromebook - this is the way!).
tools/bin/grpcurl -protoset <(tools/bin/buf build -o -) -plaintext localhost:7072 spy.v1.SpyRPCService/SubscribeSignedVAA
With parameters (using proto json encoding):
tools/bin/grpcurl -protoset <(tools/bin/buf build -o -) \
-d '{"filters": [{"emitter_filter": {"emitter_address": "574108aed69daf7e625a361864b1f74d13702f2ca56de9660e566d1d8691848d", "chain_id": "CHAIN_ID_SOLANA"}}]}' \
-plaintext localhost:7072 spy.v1.SpyRPCService/SubscribeSignedVAA
To Solana:
kubectl exec solana-devnet-0 -c setup -- client post-message Bridge1p5gheXUvJ6jGWGeCsgPKgnE3YgdGKRVCMY9o 1 confirmed ffff
To Solana as CPI instruction:
kubectl exec solana-devnet-0 -c setup -- client post-message --proxy CP1co2QMMoDPbsmV7PGcUTLFwyhgCgTXt25gLQ5LewE1 Bridge1p5gheXUvJ6jGWGeCsgPKgnE3YgdGKRVCMY9o 1 confirmed ffff
kubectl exec -it guardian-0 -- /guardiand admin send-observation-request --socket /tmp/admin.sock 1 4636d8f7593c78a5092bed13dec765cc705752653db5eb1498168c92345cd389
Locally compile protos to populate the buf cache:
make generate
Set the include path:
Guardian events can be persisted to a cloud BigTable instance by passing a GCP project and service account key to Tilt. Launch the devnet with flags supplying your database info to forward events to your cloud BigTable, rather than the local devnet BigTable emulator:
tilt up -- --num=1 --gcpProject=your-project-id --bigTableKeyPath=./your-service-account-key.json
Run the bridge UI in devnet by supplying the --bridge_ui
flag:
tilt up -- --bridge_ui
Node logs:
kubectl exec -c algod algorand-0 -- tail -f /network/Node/node.log
kubectl exec -c algod algorand-0 -- tail -f /network/Primary/node.log
Account list:
kubectl exec -c goal-kmd algorand-0 -- ./goal account list
Get yourself a working shell:
kubectl exec -c goal-kmd algorand-0 -it shell-demo -- /bin/bash
Use the --guardiand_debug
Tilt argument to run guardiand within a dlv session. The session will be exposed just like
any other Tilt services. You can then connect any IDE that supports Go debugging, like IntelliJ (add a "Go Remote"
target and specify the host and port your Tilt instance runs on).