Expose RabbitMQ for Computational clusters

[YuryHrytsuk]

RabbitMQ shall be reachable from Computational Clusters

Tasks

• [AWS] Expose rabbit for private subnets #1032
• Introduce robust / HA private DNS management
• Add configuration
  • NIH STAG
  • NIH PROD
  • AWS TIP

Related PR

• ✨Computational clusters: connect autoscaling to RabbitMQ ⚠️ osparc-simcore#7485

Useful articles

• https://docs.aws.amazon.com/Route53/latest/DeveloperGuide/best-practices-resolver-endpoint-high-availability.html
• https://stackoverflow.com/a/57323270/12124525
• https://docs.aws.amazon.com/de_de/Route53/latest/DeveloperGuide/dns-failover-configuring.html
• https://docs.aws.amazon.com/Route53/latest/DeveloperGuide/dns-failover-private-hosted-zones.html --> DNS Failover in private route53 zones

Questions

What happens if all checks are failing

If there are no healthy endpoints remaining in a resource record set, Route 53 will behave as if all health checks are passing.

Source: https://www.amazonaws.cn/en/route53/faqs/

If my healtch check turns back green, will IP Address be used again?

Yes. Manually tested by YH. Also mentioned in faqs

[YuryHrytsuk]

Idea: Return multiple IPs for a DNS Record and let clients (e.g. cURL or rabbitmq client shall try another IP from list if previos IP failed).

Not all clients (if many clients at all) will try multiple IPs. This is not a robust solution (if solution at all)

Related: https://stackoverflow.com/questions/79590113/how-does-python-requests-handles-dns-answers-with-multiple-ips

[YuryHrytsuk]

All related PRs should be merged prior to ITISFoundation/osparc-simcore#7573.

After we can close it