Moving MongoDB off Atlas without downtime
Move an Atlas cluster to MongoDB in OSStend's cloud, your cloud account or your data center, run 24/7 by AceMQ engineers.
- ✓Operated by AceMQ
- ✓160+ customers
To migrate MongoDB off Atlas with minimal downtime, MongoDB's own tool mongosync copies the data from the Atlas cluster and keeps syncing writes until you commit. mongodump and mongorestore are the offline alternative. Version limits apply: mongosync needs a MongoDB 7.0 or 8.0 destination.
Which migration method fits
mongosync is MongoDB's documented tool for moving data between clusters with minimal downtime, and an Atlas cluster can be its source. mongodump and mongorestore are simpler and work offline, so they fit when the database can stop writing for the length of the copy.
| Method | What it does | Both sides live | Source |
|---|---|---|---|
| mongosync | Replicates data and writes from one cluster to another until the sync is finalized. The source can stay active until commit. Each cluster can be a replica set or a sharded cluster. | Yes, until commit | MongoDB docs |
| mongodump and mongorestore | A binary export of a database's contents, restored into a deployment running the same major version or feature compatibility version. Writes made during the export can be included with the --oplog option. | No: writes stop for the final copy | MongoDB docs |
| Atlas snapshot download | Atlas documents downloading a snapshot and restoring it locally. Availability depends on the cluster type. | No | MongoDB docs |
Sources: MongoDB documentation, linked in each row; checked 30 Oct 2026.
How the methods were chosen: only tools that MongoDB documents are listed. MongoDB has retired mongomirror in favour of mongosync, and it was built for moves into Atlas, so it is not used here.
Version and edition limits of mongosync
mongosync only supports certain version pairs, and MongoDB says it does not test it with Community builds. The version and edition of the new cluster are therefore fixed in the design before anything is built.
| Source version | Supported destination | Note |
|---|---|---|
| 5.0 | 7.0 | 5.0 and 6.0 can only be sources. Pre-6.0 sources have extra limits, including no schema-changing writes during the migration. |
| 6.0 | 7.0 or 8.0 | Destination must be 7.0 or 8.0. |
| 7.0 | 7.0 or 8.0 | |
| 8.0 | 8.0 | mongosync does not support MongoDB 9.0. |
Source: MongoDB docs; checked 30 Oct 2026.
- Editions.
- MongoDB does not test mongosync with Community builds and, in most cases, does not offer support for it with Community deployments (MongoDB docs). Which edition the destination runs is confirmed in the design review.
- No traffic before commit.
- Until commit is called and canWrite returns true, the destination cannot take application reads or writes, and mongosync is not meant for disaster recovery or analytics copies (MongoDB docs).
- Limits are not checked for you.
- mongosync does not check compliance with its documented limitations, so the limitations page is read against the application before the sync starts (MongoDB docs).
- Atlas-only features.
- Features the application uses that exist only in Atlas, such as Atlas Search or Atlas Triggers, are listed in the review, because each needs an answer on the new side.
Sources: MongoDB documentation, linked above; checked 30 Oct 2026.
The MongoDB edition and version of the destination cluster are set in the design, with the licence terms that come with them.
The migration in phases
The new cluster is built while Atlas keeps serving the application. mongosync copies the existing data and then follows the writes, and nothing moves until the lag is small and the checks pass.
Review
Collections, indexes, data size, write rate, MongoDB version, Atlas-only features, and whether the cluster is a replica set or sharded.
Build the new cluster
Version, edition, topology and sizing come from the design, on infrastructure you control.
Sync
mongosync is started against the Atlas source and the new destination. The progress endpoint is watched until the copy is complete and the lag is low.
Verify
Document counts and sample queries are compared, and the application is tested against the new cluster with a copy of its traffic.
Cut over
Writes to Atlas stop, mongosync is committed, and the application connects to the new cluster once canWrite returns true.
The timeline and runbook for MongoDB migrations are written in the design step, from your data volume, traffic and change windows, and you approve them before the cutover is scheduled.
Cutover and rollback
The switch happens at mongosync's commit step. Atlas stays unchanged until the new cluster has taken application traffic normally, and the old cluster is retired only after that.
- The progress endpoint shows the sync caught up and the verification queries match.
- Application writes to Atlas stop, and the last writes sync.
- The sync is committed, and canWrite returns true on the new cluster.
- The application connects to the new cluster. Atlas stays available as the fallback until the new side has run normally.
Source: the mongosync documentation (MongoDB docs); checked 30 Oct 2026.
How OSStend runs the move
We run the migration and own the cutover, with both sides live until you switch. Zero downtime is the plan.
Managed MongoDB from OSStend is run 24/7 by AceMQ engineers in OSStend's cloud, your cloud account or your data center. The reasons teams leave Atlas are set out on the MongoDB Atlas alternatives page, and the migration hub lists the other moves.
MongoDB Community Server is published under the Server Side Public License (checked 31 Oct 2026). OSStend runs MongoDB in its own cloud, in your cloud account or in your data center.
AceMQ names a customer only with that customer's written permission, so most of this record stays anonymous. Ask for a similar workload in the 45-minute review.
When this move fits, and when it does not
The move fits when cost, data location or regulation is pulling the workload off Atlas. It does not when the application depends on Atlas-only features or you need a fully hosted tier today.
Contact us if
- ✓A regulated workload must run in your account or data center.
- ✓Atlas cost or data location is the reason for leaving.
- ✓You want engineers to run the sync and the cluster afterwards.
Not a fit if
- ✕The application depends heavily on Atlas-only features.
- ✕You want a free tier or self-serve sign-up. OSStend does not offer either for MongoDB.
- ✕You want to run the move yourselves: mongosync and mongodump work without OSStend.
Questions about moving off Atlas
Can MongoDB be migrated off Atlas with no downtime?
mongosync keeps syncing writes from the Atlas cluster until commit, so the application runs throughout. The switch still needs writes to stop briefly while the last changes sync (MongoDB docs, checked 30 Oct 2026).
Does mongosync work with MongoDB Community Edition?
MongoDB says it does not test mongosync with Community builds and, in most cases, does not offer support for it with Community deployments (MongoDB docs, checked 30 Oct 2026). The destination edition is chosen with that in mind.
Can mongodump be used to leave Atlas?
Yes. mongodump can export from replica sets and sharded clusters, and the files restore into a deployment on the same major version or feature compatibility version (MongoDB docs, checked 30 Oct 2026). It is an offline method, so writes stop for the final copy.
Related pages
- MongoDBMongoDB hosting: managed replica sets, shardingMongoDB hosting run 24/7 by OSStend engineers: replica sets and sharded clusters with HA, backups, point-in-time recovery, monitoring and upgrades.
- AlternativeMongoDB Atlas pricing and alternatives in 2026MongoDB Atlas pricing and alternatives: Atlas list prices, and managed MongoDB in OSStend's cloud, your cloud or your data center, with the migration run for you.
- OverviewZero-downtime migration, owned by our engineersZero-downtime migration guides for Amazon MQ, RDS, Redis, Atlas, Oracle, MySQL, Confluent and Elasticsearch, using each engine's documented method.
Plan your Atlas move with an engineer
45 minutes with an engineer: the current setup is reviewed against the documented method and the cutover is sketched. No ticket queue.
- ✓A design sketch for your workload, not a plan menu
- ✓A zero-downtime cutover path for what you run today
- ✓A price for that design, after the review