pgwire,server: clarify to SQL clients when they select the wrong tenant

[knz]

Fixes #92525.
Epic: CRDB-14537

Prior to this patch:

$ ./cockroach sql -d cluster:wo
ERROR: server closed the connection.
Is this a CockroachDB node?
unexpected EOF

After this patch:

$ ./cockroach sql -d cluster:woo
ERROR: service unavailable for target tenant (woo)
SQLSTATE: 08000
HINT: Double check your "-ccluster=" connection parameter or your "cluster:" database name prefix.

Release note: None

[cockroach-teamcity]

This change is 

[rafiss]

nice improvement! had one question

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @ecwall, @knz, and @stevendanna)

---

pkg/sql/pgwire/pre_serve.go line 215 at r1 (raw file):

		pgerror.Newf(pgcode.ConnectionException,
			"service unavailable for target tenant (%v)", tenantName),
		`Double check your "-ccluster=" connection option or your "cluster:" database name prefix.`)

would it be possible to get this error is someone were using our SNI routing feature? if so, this hint would be confusing, as they are not expected to provide a -c cluster option in that case

[knz]

I have drafted the following answer to your question:

at this point, no CC serverless client would ever see the new error, because those connections would never be routed to a host/storage node.
in the future, we may want to support SNI to route SQL clients to 1 particular tenant out of many "hosted" using shared-process multitenancy on the same process. When that time comes, we can extend the error message to mention SNI.

now asking for serverless friends for confirmation

[knz]

@rafiss here's the response from Jeff - WDYT?

at this point, no CC serverless client would ever see the new error, because those connections would never be routed to a host/storage node.

This is correct. If network level security is working correctly there should be no way for a Serverless user to reach the storage servers with a sql connection. If they are able to reach the storage servers.

in the future, we may want to support SNI to route SQL clients to 1 particular tenant out of many “hosted” using shared-process multitenancy on the same process. When that time comes, we can extend the error message to mention SNI.

I see a future where this is useful for non-serverless clusters, but I don’t think Serverless would ever offer this in CC. If it were an option we may use it to simplify certain maintenance tasks.

[rafiss]

thanks for checking!

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @ecwall, @knz, and @stevendanna)

[knz]

cheers!

bors r=rafiss

[craig]

This PR was included in a batch that was canceled, it will be automatically retried

[craig]

Build failed (retrying...):

• Bazel Essential CI (Cockroach)

[craig]

Build succeeded:

• Bazel Essential CI (Cockroach)