Add dynamic ingress hostname and NodePort env vars

[pcfreak30]

Hello.

So this PR might be over-complicated/over engineered. I have not messed with k8 and spent several days it feels figuring this out, and had a lot of AI help.

The key things it adds are

ingressHost := pmanifest.IngressHost(lid, b.Name())
	env = addIfNotPresent(envVarsAlreadyAdded, env, envVarAkashIngressHostname, fmt.Sprintf("%s.%s", ingressHost, b.settings.DeploymentIngressDomain))

	svc := &b.deployment.ManifestGroup().Services[b.serviceIdx]

	// Add hostnames from service expose configurations
	for _, expose := range svc.Expose {
		if expose.IsIngress() {
			// Add custom hostnames if specified
			for idx, hostname := range expose.Hosts {
				env = addIfNotPresent(envVarsAlreadyAdded, env,
					fmt.Sprintf("%s_%d_%d", envVarAkashIngressCustomHostname, expose.Port, idx),
					hostname)
			}
		}

		if expose.Global {
			// Add external port mappings
			if svc := b.service; svc != nil {
				if nodePort, exists := svc.portMap[int32(expose.Port)]; exists {
					env = addIfNotPresent(envVarsAlreadyAdded, env,
						fmt.Sprintf("%s_%d", envVarAkashExternalPort, expose.Port),
						fmt.Sprintf("%d", nodePort))
				}
			}
		}
	}

Now the hostnames are much easier to deal with overall. However the ports put me through rabbit hole hell.

• I am not sure if a service can or should be moved to be created before a statefulset/deployment.
• Ports are only assigned AFTER creation of the service, so we have to run service apply twice.
• We have to ALSO re-apply the statefulset/deployment to use the generated env vars.
• We have to do a hack on the builder factories due to the weird struct embedding of workload? So we pass an optional service pointer to the stateful/deployment builder factory and have it call a new setService helper. Found also it has to be called direct and not on the workload reference, using go's pass through on struct embedding.
• waitForProcessedVersion was created along with k8 retry helper imported because I was concerned about cached data being returned as it seems some stuff is processed async, and i got a conflict error at-least 1 time.

This PR is currently a draft to get feedback on how this should be best implemented vs what I have done. I also generally don't do unit tests in most cases so if they are needed, will need to be advised where at.

This is what has currently worked for me and I will be deploying this to prod for my projects workloads.

The use case for me is I have mysql master & slave servers that has to self identify themselves (host and port) and register with etcd. I have a docker image im building that acts as one side of the cluster at https://github.com/LumeWeb/akash-mysql with https://github.com/LumeWeb/akash-proxysql being the other.

Kudos.

[troian]

@pcfreak30 thanks for bringing this up.
can you first create an issue in https://github.com/akash-network/support and describe issue and proposed solution

[pcfreak30]

@pcfreak30 thanks for bringing this up. can you first create an issue in https://github.com/akash-network/support and describe issue and proposed solution

I can, however, pretty much everything needed has already been said here? Please clarify?