weaveworks-liquidmetal / microvm-operator Goto Github PK

This is currently configurable at the wrong level

changing the replicas at the deployment level currently has no effect

Right now you can only create a microvm on a single host.

It would be good to say "take this spec, create it on these hosts".

Note that this is not about us detecting whether the hosts have capacity, or trying to achieve an even spread of microvms across a pool. This is not "i want to create 10 mvms of the same spec across 3 hosts, figure it out. i expect to end up with 10 total".

This is "i want to create 10 mvms of the same spec on each of these 3 hosts. i expect to end up with 30 total".

If we would rather have the other option (either instead of or as well as), we can create a new ticket.

The two repos have a lot in common, eg:

• The service
• Some bits of spec
• Some scopes
• The client

Create some sensible interfaces and move all these into another repo (or several other repos) and use that instead. Then go update capmvm.

start with:

kind: MicrovmDeployment
spec:
  hosts:
  - name: host1
    endpoint: 127.0.0.1:9090

create.

update to:

kind: MicrovmDeployment
spec:
  hosts:
  - name: host2
    endpoint: 127.0.0.1:9091

I should see the RS deleted from host1 and a new one created on host2. Right now nothing has noticed that this has changed

I left this out of the replicaset for now but it would be useful to have.

Any microvm would be deleted anyway in the recreate, but it saves a command.

Right now you can only create one microvm at a time.

It would be good to create as many as you like with the same spec. MicrovmReplicaSet or something.

apiVersion: infrastructure.flintlock.x-k8s.io/v1alpha1
kind: MicrovmReplicaSet
metadata:
  name: foobar
spec:
  replicas: 3
  host:
    name: host1
    endpoint: 1.2.3.4:9090
  template:
    spec:
      sshPublicKeys:
      - user: "root"
        authorizedKeys:
        - "ssh-ed25519 foobar"
      userdata: |
        #!/bin/bash
        echo "hi from my microvm!"
      kernel:
        filename: boot/vmlinux
        image: ghcr.io/weaveworks-liquidmetal/flintlock-kernel:5.10.77
      kernelCmdline: {}
      memoryMb: 2048
      vcpu: 2
      networkInterfaces:
      - guestDeviceName: eth1
        type: macvtap
      rootVolume:
        id: root
        image: ghcr.io/weaveworks-liquidmetal/capmvm-kubernetes:1.21.8

LET'S HAVE SOME OF THOSE YEH?