Machine constraint to target a cluster member when deploying to LXD

[Simon Déziel]

When launching instances on a LXD cluster, one can either let the cluster decide on the instance's placement or request a specific cluster member to be the target. Here is how to do the explicit targeting with `lxc`:

```
lxc launch images:ubuntu/jammy my-lxd-cluster: --target nodeA
```

Having a Juju machine constraint to accomplish the same would be useful.

[Stéphane Graber (stgraber)]

Not a constraint but you can use the availability zones to achieve this.

root@juju01:~# juju add-machine zone=cluster05
created machine 3
root@juju01:~# juju status
Model Controller Cloud/Region Version SLA Timestamp
default nsec-default nsec/default 2.9.29 unsupported 17:32:00Z

Machine State DNS Inst id Series AZ Message
3 pending 240.28.0.167 juju-378e3e-3 focal Running

root@cluster01:~# lxc list juju-378e3e-3
+---------------+---------+---------------------+------+-----------+-----------+-----------+
| NAME | STATE | IPV4 | IPV6 | TYPE | SNAPSHOTS | LOCATION |
+---------------+---------+---------------------+------+-----------+-----------+-----------+
| juju-378e3e-3 | RUNNING | 240.28.0.167 (eth0) | | CONTAINER | 0 | cluster05 |
+---------------+---------+---------------------+------+-----------+-----------+-----------+

[Stéphane Graber (stgraber)]

To clarify, a `zone=NAME` during `add-machine` works properly, however one cannot use the `zones` contraint. Juju accepts it but then completely disregards it and deploys anywhere.

[Stéphane Graber (stgraber)]

```
root@juju01:~# juju deploy lxd -n 3 --config mode=cluster --constraints zones=cluster03,cluster03,cluster03
Located charm "lxd" in charm-hub, revision 168
Deploying "lxd" from charm-hub charm "lxd", revision 168 in channel stable on focal
```

```
controller-0: 18:25:56 INFO juju.worker.provisioner provisioning in zones: [cluster01 cluster02 cluster03 cluster04 cluster05 cluster06]
controller-0: 18:25:56 INFO juju.worker.provisioner found machine pending provisioning id:11, details:11
controller-0: 18:25:57 INFO juju.worker.provisioner provisioning in zones: [cluster01 cluster02 cluster03 cluster04 cluster05 cluster06]
controller-0: 18:25:57 INFO juju.worker.provisioner found machine pending provisioning id:12, details:12
controller-0: 18:25:57 INFO juju.worker.provisioner trying machine 11 StartInstance in availability zone cluster03
controller-0: 18:25:58 INFO juju.worker.provisioner provisioning in zones: [cluster01 cluster02 cluster03 cluster04 cluster05 cluster06]
controller-0: 18:25:58 INFO juju.worker.provisioner found machine pending provisioning id:13, details:13
controller-0: 18:25:58 INFO juju.worker.provisioner trying machine 12 StartInstance in availability zone cluster03
controller-0: 18:25:58 INFO juju.worker.provisioner trying machine 13 StartInstance in availability zone cluster03
controller-0: 18:26:06 INFO juju.worker.provisioner started machine 11 as instance juju-378e3e-11 with hardware "arch=amd64 cores=0 mem=0M", network config [], volumes [], volume attachments map[], subnets to zones [], lxd profiles []
controller-0: 18:26:07 INFO juju.worker.provisioner started machine 12 as instance juju-378e3e-12 with hardware "arch=amd64 cores=0 mem=0M", network config [], volumes [], volume attachments map[], subnets to zones [], lxd profiles []
controller-0: 18:26:08 INFO juju.worker.instancemutater.environ no changes necessary to machine-12 lxd profiles ([default juju-default])
controller-0: 18:26:08 INFO juju.worker.instancemutater.environ no changes necessary to machine-11 lxd profiles ([default juju-default])
controller-0: 18:26:09 INFO juju.worker.instancepoller machine "11" (instance ID "juju-378e3e-11") instance status changed from {"running" "Container started"} to {"running" "Running"}
controller-0: 18:26:09 INFO juju.worker.instancepoller machine "11" (instance ID "juju-378e3e-11") has new addresses: [local-fan:240.12.0.139@alpha]
controller-0: 18:26:09 INFO juju.worker.instancepoller machine "12" (instance ID "juju-378e3e-12") instance status changed from {"running" "Container started"} to {"running" "Running"}
controller-0: 18:26:10 INFO juju.worker.instancepoller machine "12" (instance ID "juju-378e3e-12") has new addresses: [local-fan:240.17.0.11@alpha]
controller-0: 18:26:38 INFO juju.worker.provisioner started machine 13 as instance juju-378e3e-13 with hardware "arch=amd64 cores=0 mem=0M", network config [], volumes [], volume attachments map[], subnets to zones [], lxd profiles []
controller-0: 18:26:43 INFO juju.worker.instancemutater.environ no changes necessary to machine-13 lxd profiles ([default juju-default])
controller-0: 18:26:46 INFO juju.worker.instancepoller machine "13" (instance ID "juju-378e3e-...

[Vitaly Antonenko (anvial)]

This is potentially an interesting constraint for the machine and could be discussed as a feature.