wasmCloud Application Deployment Manager (wadm)

The wasmCloud Application Deployment Manager (wadm) enables declarative wasmCloud applications. It's responsible for managing a set of application deployment specifications, monitoring the current state of an entire lattice, and issuing the appropriate lattice control commands required to close the gap between observed and desired state.

Using wadm

Install & Run

You can easily run wadm by downloading wash and launching it alongside NATS and wasmCloud. Then, you can use the wash app command to query, create, and deploy applications.

wash up -d    # Start NATS, wasmCloud, and wadm in the background
wash app list # Query the list of applications

If you prefer to run wadm separately and/or connect to running wasmCloud hosts, you can instead opt for using the latest GitHub release artifact and executing the binary. Simply replace the latest version, your operating system, and architecture below. Please note that wadm requires a wasmCloud host version >=0.63.0

# Install wadm
curl -fLO https://github.com/wasmCloud/wadm/releases/download/<version>/wadm-<version>-<os>-<arch>.tar.gz
tar -xvf wadm-<version>-<os>-<arch>.tar.gz
cd wadm-<version>-<os>-<arch>
./wadm

Deploying an application

Take the following manifest and save it locally (you can also download this from echo.yaml):

apiVersion: core.oam.dev/v1beta1
kind: Application
metadata:
  name: echo
  annotations:
    version: v0.0.1
    description: "This is my app"
spec:
  components:
    - name: echo
      type: component
      properties:
        image: wasmcloud.azurecr.io/echo:0.3.7
      traits:
        - type: spreadscaler
          properties:
            instances: 1
        - type: linkdef
          properties:
            target: httpserver
            values:
              address: 0.0.0.0:8080

    - name: httpserver
      type: capability
      properties:
        contract: wasmcloud:httpserver
        image: wasmcloud.azurecr.io/httpserver:0.17.0
      traits:
        - type: spreadscaler
          properties:
            instances: 1

Then, use wadm to put the manifest and deploy it.

wash app put ./echo.yaml
wash app deploy echo

🎉 You've just launched your first application with wadm! Try curl localhost:8080/wadm and see the response from the echo WebAssembly module.

When you're done, you can use wadm to undeploy the application.

wash app undeploy echo

Modifying applications

wadm supports upgrading applications by putting new versions of manifests and then deploying them. Try changing the manifest you created above by updating the number of echo instances.

<<ELIDED>>
  name: echo
  annotations:
    version: v0.0.2 # Note the changed version
    description: "wasmCloud echo Example"
spec:
  components:
    - name: echo
      type: component
      properties:
        image: wasmcloud.azurecr.io/echo:0.3.5
      traits:
        - type: spreadscaler
          properties:
            instances: 10 # Let's run 10!
<<ELIDED>>

Then, simply deploy the new version:

wash app put ./echo.yaml
wash app deploy echo v0.0.2

If you navigate to the wasmCloud dashboard, you'll see that you now have 10 instances of the echo actor.

Documentation for configuring the spreadscaler to spread actors and providers across multiple hosts in a lattice is forthcoming.

Responsibilities

wadm has a very small set of responsibilities, which actually contributes to its power.

• Manage Application Specifications - Manage models consisting of desired state. This includes the creation and deletion and rollback of models to previous versions. Application specifications are defined using the Open Application Model. For more information on wadm's specific OAM features, see our OAM README.
• Observe State - Monitor wasmCloud CloudEvents from all hosts in a lattice to build the current state.
• Take Compensating Actions - When indicated, issue commands to the lattice control interface to bring about the changes necessary to make the desired and observed state match.

🚧 Advanced

In advanced use cases, wadm is also capable of:

• Monitoring multiple lattices.
• Running multiple instances to distribute load among multiple processes, or for a high-availability architecture.

🚧 The above functionality is somewhat tested, but not as rigorously as a single instance monitoring a single lattice. Proceed with caution while we do further testing.

API

Interacting with wadm is done over NATS on the root topic wadm.api.{prefix} where prefix is the lattice namespace prefix. For more information on this API, please consult the wadm Reference.

Known Issues/Missing functionality

As this is a new project there are some things we know are missing or buggy. A non-exhaustive list of these can be found below:

• It is technically possible as things stand right now for a race condition with manifests when a manifest is updated/created and deleted simultaneously. In this case, one of the operations will win and you will end up with a manifest that still exists after you delete it or a manifest that does not exist after you create it. This is a very unlikely scenario as only one person or process is interacting with a specific, but it is possible. If this becomes a problem for you, please let us know and we will consider additional ways of how we can address it.
• Manifest validation is not yet implemented. Right now wadm will accept the manifest blindly so long as it can parse it. It will not validate that the model name is valid or if you specified entirely invalid properties. This will be added in a future version.
• Nondestructive (e.g. orphaning resources) undeploys are not currently implemented. You can set the field in the request, but it won't do anything
• If wadm discovers a provider in the lattice that isn't already started via a manifest, the manifest won't be able to reconcile until another start provider event is received. This will require a feature add to the ctl client to fix
• All manifests belonging to a lattice must be using the same version of actors and providers. This is an important limitation of the host RPC protocol. We are open to ideas about how to best handle manifests with different actor/provider versions
• When running multiple wadm processes, you can still get a little bit of jitter with starting actors and providers (in some cases). This will always resolve after a few ticks and isn't a huge problem as actors are "cheap" from a compute standpoint. If anyone would like to submit a PR to make this better, we'd love to see it!

References

The wasmCloud Application Deployment Manager (wadm) originally came from the autonomous lattice controller RFC and then was reworked in version 0.4 in response to the tidying and productioning RFC.

• Autonomous Agents
• Autonomous Agent (Wikipedia)
• Control Loop
• Hashicorp Vault