Terminate SSL connections in pgsrv proxy

[scsmithr]

We currently don't support SSL. Specifically, this issue is to support sslmode=require in the connection string.

[scsmithr]

Might be dependent on how we structure the rest of the extended query protocol.

[greyscaled]

This one is going to block first customers. Not necessary in an alpha but 100% blocks moving to a "beta" or even bill-able offering IMO.

I don't have data to back up this statement, just experience and spidey sense.

LMK if disagree of course!

[greyscaled]

I think priority can stay medium since it's not blocking alpha users. But I want to keep in mind/track it for when we're ready to start seriously implementing the earliest versions of billing and external trial users/potential customers.

I think a few things can proceed this:

• A design doc for billing
• An early implementation of billing that doesn't actually bill, but --dry-runs it
• Refactoring pgsrv to use cloud client etc
• Adding missing (critical) tests to pgsrv + health check

I think once we accomplish those things, we're getting very close to a bill-able/beta state, and we should implement this

[scsmithr]

Agreed. Happens to also block setting up airbyte (easily).

[greyscaled]

Yea, we're going to want external users to be able to connect with whatever tools they prefer. It's OK if we constrain earliest versions of alpha to be terminal and dashboard only.

In fact, I'll add that constraint to December MVP.

[scsmithr]

It's OK if we constrain earliest versions of alpha to be terminal and dashboard only.

There's really two user-facing areas of focus for us: transactional and analytical workloads. If we take someone like Nathan, the first thing he did was try to run migrations for a transactional service. This is going to be very hard for us to fully support short term.

But if we suggest hooking up GlareDB as a destination in Airbyte, then we can get up and running on analytical workloads right away. In Airbyte's case, all they really need is to be able to create tables and run inserts. Then the analytics is really just a bunch of selects. These are operations that we support right now (whether or not it's performant is TBD).

I think for running the actual select queries, the terminal and dashboard for alpha is fine. But I think we'll need to make ingesting data a priority sooner rather than later (both for getting alpha users up and running with a database they can actually mess around with, as well as for internal testing). All this to say I think getting Airbyte ingestion unblocked is very important right now, so I think this issue is high prio.

[greyscaled]

I'm definitely swayed by that. I think the epic (which contains this ticket) #366 is high priority - as in I now realize we don't need to get closer to a billing state, but we need pgsrv to have an implementation we can test and depend on, including this implementation.

Will udpate accordingly. I'm going to consider this soft-blocked by some refactors in pgsrv.

Moved to backlog, high priority.

[greyscaled]

whether or not it's performant is TBD

See: #364

[greyscaled]

This is done