Added example of Versioned PiWind deployments

[sambles]

Updated deployment for PR OasisLMF/OasisPlatform#931

This adds two deployments piwind, one for each API version v1 and v2. There is currently an issue with auto-scaling between both these deployments. For testing disable the worker-controller and manually scale the piwind deployments.

$ kubectl scale --replicas=0 deployment/oasis-worker-controller
deployment.apps/oasis-worker-controller scaled

$ kubectl scale --replicas=1 deployment/worker-oasislmf-piwind-1-v1
deployment.apps/worker-oasislmf-piwind-1-v1 scaled
$ kubectl scale --replicas=1 deployment/worker-oasislmf-piwind-1-v2
deployment.apps/worker-oasislmf-piwind-1-v2 scaled

To edit the worker v1 image switch the tag in these lines, for example dev -> 1.28.5 and rerun ./deploy.sh models

OasisAzureDeployment/settings/helm/workers/oasislmf-piwind-1.yaml

Lines 2 to 8 in 4fd5572

	oasislmf_piwind_1: # A name that is unique among all workers
	supplierId: OasisLMF # Must be identical to supplier in the model data file share
	modelId: PiWind # Must be identical to name in the model data file share
	modelVersionId: "1" # Must be identical to version in the model data file share
	apiVersion: "v1" # Single Server execution
	image: ${ACR}/coreoasis/model_worker # The path to your image, ${ACR} will automatically be replaced with your environments URL
	version: dev # Version tag of your image

Troubleshooting - Celery DB connection error

If v1 workers are crashing with the following error

kubectl logs worker-oasislmf-piwind-1-v1-56977d8c49-q8t6m
Defaulted container "worker" out of: worker, init-tcp-wait-by-secret (init)
wait-for-it.sh: waiting 60 seconds for broker:5672
wait-for-it.sh: broker:5672 is available after 0 seconds
wait-for-it.sh: waiting 60 seconds for oasis-j25vz4tnmuh5m.privatelink.postgres.database.azure.com:5432
wait-for-it.sh: oasis-j25vz4tnmuh5m.privatelink.postgres.database.azure.com:5432 is available after 0 seconds
[2023-12-14 13:35:46,296: CRITICAL/MainProcess] Unrecoverable error: ValueError("Port could not be cast to integer value as '******' ")

The error is caused by special characters in celery password, tripping up the celery DB connection. Note that v2 don't have this issue because they are escaped within the celery app.

The easiest workaround is updating the celery password by deleting the celery-db-password entry from the Key Vault.

Then rerunning the deployment ./deploy.sh base using this branch, this generates a new password without the problematic characters.