Skip to content
TCP proxy peer address and Hobby image retention — read-only clarification
andreslink74
OP

18 days ago

Please clarify two platform behaviors before any networking or deployment change. This request authorizes no changes.

  1. For a public TCP Proxy to PostgreSQL, what peer address does the server see? Is the original client IP preserved, replaced by a Railway proxy address, or can the connection ever appear as 127.0.0.1 / ::1? We need an authoritative statement about whether external traffic can match loopback-only pg_hba.conf rules. Does Railway guarantee a non-loopback peer, and is this documented? No proxy should be enabled to answer this.

  2. The Hobby image-retention documentation says 72 hours after removal and that Redeploy rebuilds outside retention. With a removed deployment older than 72 hours, canRedeploy=true and canRollback=false, is reuse of the original image without rebuilding supported? Is there a read-only way to verify retained image availability without invoking deploymentRedeploy?

Please distinguish general documented guarantees from anything requiring a live test. No logs, secrets, application records or infrastructure changes are needed. Service N/A: platform-level questions only.

Solved

1 Reply

Railway
BOT

18 days ago

On the TCP Proxy, the proxy opens its own connection to your service and relays the bytes. PostgreSQL sees the proxy as the peer, not the original client. The client IP is not passed through, and no PROXY protocol header is added. The peer addresses you may see are not a published or stable range, so don't allowlist them in pg_hba.conf. We can't give you a documented guarantee about which address family or range the peer will come from. Treat every TCP Proxy connection as external traffic and require authentication for it, rather than counting on loopback rules never matching.

On image retention, Hobby keeps an image for 72 hours after its deployment is removed. Rolling back within that window reuses the original image. After the window, rollback isn't offered, and Redeploy rebuilds from the deployment's original source and variables. So reusing the original image without a rebuild is not supported once retention has passed, which is consistent with canRollback=false. canRollback is the read-only signal for whether rollback is available. It doesn't confirm that a particular image file has been physically deleted, because cleanup after expiry happens automatically on its own schedule. See the image retention policy.

Everything above is documented behaviour. None of it needs a live test.


Status changed to Awaiting User Response Railway • 18 days ago


Railway
BOT

11 days ago

This thread has been marked as solved automatically due to a lack of recent activity. Please re-open this thread or create a new one if you require further assistance. Thank you!

Status changed to Solved Railway • 11 days ago


Welcome!

Sign in to your Railway account to join the conversation.

Loading...