FAQ: Storage
On this page
What happens when I reach my Atlas storage limit?
The result of reaching your Atlas storage limit depends on the Atlas cluster you are using.
- For shared clusters (
M0,M2,M5), the maximum storage is a hard limit and cannot be exceeded. You can add additional storage by upgrading to a dedicated cluster (M10+). For details on how Atlas calculates storage limits for shared clusters, see this section of the FAQ. - By default,
M10+clusters auto-expand storage based on disk usage thresholds. To modify this setting to a fixed storage limit, refer to the Modify a Cluster page.
If you attempt to write to a shared cluster that does not have space for the desired write operation, Atlas displays an error message similar to the following:
WriteResult({ "writeError": { "code": 8000, "errmsg": "you are over your space quota, using 513 MB of 512 MB" } })
To learn about the differences between shared and dedicated clusters, see Atlas M0 free cluster, M2, and M5 Limitations.
You can configure alerts which trigger once your allocated storage reaches a specified threshold. Atlas calculates allocated storage using metrics returned by the dbStats() command. To learn more about storage alerts, see DB Storage alert conditions.
How does Atlas calculate storage limits for shared clusters (M0, M2, M5)
Atlas calculates the storage limit for shared clusters based on
data usage, as opposed to the storageSize metric used by
non-shared clusters (which includes compression). Atlas determines
data usage by summing a cluster's dataSize and indexSize. You
can issue the db.stats()
method to view the values of these fields.
Without using Provisioned IOPS on MongoDB Atlas on AWS, what kind of IOPS should I expect?
Atlas provides an estimate of how many 16K IOPS you can expect, calculated as the lesser of 3 IOPS per provisioned GB, or the cluster node's maximum IOPS capacity.
However, non-provisioned IOPS on AWS can burst above this estimate or drop below it. As a result, customers who are interested in consistent IOPS throughput should consider leveraging Provisioned IOPS.