# Cloud SQL connectors can now follow SQL Server primary failover through one write endpoint

Google Cloud SQL for SQL Server now lets the Auth Proxy and language connectors use a write-endpoint DNS name that follows the writable primary after failover or switchover. The endpoint already existed; the new connector support removes another piece of application-side primary discovery.

Cloud SQL’s SQL Server HA path is becoming more transparent to applications: supported proxies and connectors can target one write endpoint and be redirected when the primary changes. Teams still need retry-safe connection handling around the failover itself.

- Status: Active
- Published: 2026-09-05T17:16:56+12:00
- Updated: 2026-09-05T17:16:56+12:00
- Categories: Cloud & Infrastructure, Cloud Platforms, Databases & Storage
- Tags: Cloud SQL, Cloud SQL Auth Proxy, database failover, Google Cloud, high availability, SQL Server
- Canonical HTML: https://beyondthe.news/dossiers/cloud-sql-sql-server-write-endpoint-connectors-failover

## What changed

On September 4, Google added support for connecting Cloud SQL for SQL Server applications to a write endpoint through the Cloud SQL Auth Proxy and supported language connectors. A write endpoint is a DNS name associated with the writable primary. When a replica failover or switchover changes which instance is primary, Google updates the endpoint so new connections are directed to the new primary. Write endpoints were already available; the material change is that applications using Google’s supported connector/proxy layer can now use that abstraction directly instead of separately discovering or pinning an instance address.

## Why it matters

High-availability databases only help if applications can find the new primary reliably. Hard-coded instance endpoints create an extra failure mode during switchover and make connection logic more complicated. Connector support makes the write endpoint a more complete application boundary for Cloud SQL SQL Server. It does not make failover invisible: existing connections can still break and applications need sensible retry behavior, but it reduces the amount of topology knowledge the application must carry.

## The connector path can now target the role, not the instance

Cloud SQL Auth Proxy and supported language connectors can be configured with the write-endpoint DNS name. That shifts the connection target from a specific database instance to the current writable role.

## Failover changes the primary behind the endpoint

When Cloud SQL promotes a replica during failover or a planned switchover, the write endpoint is redirected to the new primary. New connections using that name can therefore follow the topology change without an application configuration rewrite.

## This does not remove normal database failover behavior

A DNS-backed endpoint does not preserve an already-open connection through a primary transition. Applications should still expect transient connection failures and use retry/backoff patterns appropriate to their transaction semantics.

## Key details

- The change applies to Cloud SQL for SQL Server.
- Cloud SQL Auth Proxy can now connect through a write endpoint.
- Cloud SQL language connectors can also use the write endpoint.
- The endpoint tracks the writable primary after failover or switchover.
- Write endpoints existed before this release; connector/proxy support is the new capability.
- Applications still need to handle interrupted connections during failover.

## Builder takeaways

- Replace hard-coded primary instance addresses with the write endpoint where your Cloud SQL SQL Server architecture supports it.
- Test planned switchover with your actual proxy/connector and connection pool before relying on the abstraction in production.
- Make transaction retries idempotent where possible; following the new primary does not tell the application whether an interrupted transaction committed.
- Avoid caching resolved addresses longer than the supported connection stack expects during failover events.

## What to watch

- Connector-specific guidance on reconnection timing during failover.
- Expansion of the same write-endpoint integration to other Cloud SQL engines or client paths.
- Operational data on DNS/update convergence during planned and unplanned transitions.

## Uncertainties

- The feature simplifies primary discovery but does not guarantee zero-downtime sessions.
- Failover experience still depends on application connection pools, DNS behavior and retry logic.

## Sources

- [Cloud SQL for SQL Server release notes](https://docs.cloud.google.com/sql/docs/sqlserver/release-notes) — Google Cloud · primary · 2026-09-04T00:00:00+12:00. Primary announcement that Auth Proxy and language connectors can connect using write endpoints.
- [Connect using a write endpoint](https://docs.cloud.google.com/sql/docs/sqlserver/connect-write-endpoint) — Google Cloud · primary documentation. Current behavior of write endpoints across primary failover and switchover.