MongoDB

Overview

The MongoDB connector opens a change stream on the collections you choose, so workflows react to writes as they commit. Each change arrives as a typed event with the full document attached, and every event shows up in the trace of the run it started.

Teams use it to refund a Stripe charge when an order document is cancelled, to open a Zendesk ticket when a device reports an error and to forward product events to Kafka for the data team.

Supported versions

• MongoDB 5.0 to 8.0, as a replica set or sharded cluster

• MongoDB Atlas dedicated clusters and self-managed deployments

• Connections over TLS with SCRAM or X.509, plus Atlas private endpoints

How change streams work

Tessel watches the database with a filter for your collections and asks for the current document on updates. Each event carries a resume token. Tessel stores the token after every delivery, so after a restart it resumes from the next change instead of starting over.

db.watch(
  [{ $match: { "ns.coll": { $in: ["orders", "customers"] } } }],
  { fullDocument: "updateLookup" }
)

db.watch(
  [{ $match: { "ns.coll": { $in: ["orders", "customers"] } } }],
  { fullDocument: "updateLookup" }
)

db.watch(
  [{ $match: { "ns.coll": { $in: ["orders", "customers"] } } }],
  { fullDocument: "updateLookup" }
)

Resume window

A resume token is only valid while its change is still in the oplog. Size the oplog for at least 24 hours of writes, or set a minimum retention period on Atlas. If a token has aged out, Tessel takes a fresh snapshot and notes it in the trace.

Permissions needed

Create a database user for Tessel with a custom role. It needs the changeStream action and read access on the watched collections. Write actions are only required if a workflow inserts or updates documents.

• changeStream and find on each watched collection

• The built-in read role works for a quick start

• insert and update only on collections you write back to