Add Support for MultiSubnetFailover when using a High Availability Group

[pip8786]

Currently Prisma doesn't support the MultiSubnetFailover option of SQL Server. See prisma/prisma#9187. I think this would require support at the driver level so I'm adding an issue here.

Other info: https://learn.microsoft.com/en-us/dotnet/framework/data/adonet/sql/sqlclient-support-for-high-availability-disaster-recovery

[dariusj18]

An example of implementation is here:

https://github.com/tediousjs/tedious/blob/a51e57dd20f7739e1b2cb21600ae56fe93a4ce96/src/connection.ts#L2067

connectInParallel

https://github.com/tediousjs/tedious/blob/a51e57dd20f7739e1b2cb21600ae56fe93a4ce96/src/connector.ts#L9

[pip8786]

Thanks @dariusj18 . That seems relatively simple so maybe the maintainers could implement this pretty quickly in their rust code. I tried looking at the connection code and I'm definitely not qualified to try a PR in Rust.

[dariusj18]

I'd give it stab, but I am kinda worried that PRs won't be merged in. There hasn't been much activity and a few people that have commented say things like "I'm not the owner of this crate anymore"

So I don't know if anyone is really maintaining it.

[dariusj18]

Upon further further evaluation, the TcpStream connection is created in a few places

tiberius/src/sql_browser/async_std.rs

Line 16 in e74b556

async fn connect_named(builder: &crate::client::Config) -> crate::Result<Self> {

tiberius/src/sql_browser/tokio.rs

Line 18 in e74b556

async fn connect_named(builder: &Config) -> crate::Result<Self> {

tiberius/src/sql_browser/smol.rs

Line 17 in e74b556

async fn connect_named(builder: &Config) -> crate::Result<Self> {

it seems that these exist for different library compatibilities.

[pip8786]

So do you think we need to post an issue on prisma-engines as well or is this still the right place?

[dariusj18]

This is the right place.

[tylerclendenin]

I created a simple example PR #357 with changes I assume would be needed. This is the first time I have touched rust, so I am not sure if it is correct. Also, if it is correct, the same changes would be needed to be made to tokio and smol implementations.

[cc8dea2e-4a29-4394-af6d-af0ccc73647a]

This is a really important property, and this is the only crate where I found a mention of MultiSubnetFailover. Unfortunately, I get the impression that the project is no longer being developed. 😢

[Robert-OP]

Hello! Thank you for looking into this. I think it's related, we using Prisma and are trying to connect MSSQL database instance inside a Cluster and we need to set MultiSubnetFailover, otherwise we are getting timeout on our connections. We have also increased the poolTimeout argument, but without success.

PrismaClientInitializationError: Timed out fetching a new connection from the connection pool. 
More info: http://pris.ly/d/connection-pool (Current connection pool timeout: 10, connection limit: 65)
at r (file:///app/node_modules/@prisma/client/runtime/library.mjs:119:2565)
error Command failed with exit code 1.

[srkaj]

@Robert-OP , could you try the srkaj:multi-subnet-failover branch referenced in #378 ? I think it is working and reasonably complete, but testing from others than myself would be good.

[cc8dea2e-4a29-4394-af6d-af0ccc73647a]

@srkaj i tried the srkaj:multi-subnet-failover branch. It works for me, thx!