A client of ours want to deploy the TSA in its own cluster to avoid full mesh between all the MTA instances. We saw in Kumo documentation, it is said to configure the MTA to publish data for all TSA instances, but they want to deploy it as a scalable ecs task (fargate). Any suggestions?
What does your client means by “avoid full mesh”?
By Kumo documentation, all MTA instances need to publish to all TSA instances, and by default run the TSA daemon alongside the MTA which cause full mesh network.
When you scale up your MTA cluster, Kumo recommended to create a dedicated cluster for TSA - From client
So build that client a dedicated cluster?
This sounds like another “Fargate” question. Email is inherently stateful and the data in the TSA database also needs to be persistent, so using Fargate is probably not a great solution.
It might be worth asking “Why Fargate”? I suspect the answer is cost, but I believe you will spend more on external scaffolding to supplement Fargate than Fargate will save you.
Thanks Tom