Oplog sizing for Percona Link for MongoDB¶

Percona Link for MongoDB synchronizes data between MongoDB replica sets using change streams. For the sync to complete successfully, the required operations must remain in the source’s oplog until they are applied. If operations expire from the oplog before being processed, the sync will fail.

Source cluster oplog requirements¶

After the initial data copy, PLM applies ongoing changes using the oplog from the source cluster. If the oplog window is too short, PLM may fall behind and lose access to required changes. To avoid this, increase the oplog retention window using the replSetResizeOplog command.

This is especially important when:

migrating large datasets
pausing the sync
experiencing slow sync performance

Destination storage considerations¶

The destination cluster must have enough space to store both the full dataset and the oplog entries that represent applied changes. To reduce the disk usage on the destination, you can:

Lower the oplog retention window using the storage.oplogMinRetentionHours configuration option in MongoDB
Keep the oplog size minimal if no extended retention is needed. You can manage the oplog size using the replication.oplogSizeMB option.

Track the data sync progress¶

During the initial data clone, PLM clones data and then applies oplog entries on top. You track the sync progress and fine-tune PLM. Here’s how:

Extend the oplog window if the lag approaches its limit¶

Evaluate the oplog size required for initial data clone with the following command:
```
> db.getReplicationInfo().timeDiff
```
The value you get is the minimum oplog window, in seconds.
Compare this value to the current sync lag using the PLM status command and the lagTime field. If the lagTime approaches the oplog window, extend the window using the replSetResizeOplog with a higher minRetentionHours value.

Improve the sync performance to reduce lag¶

If the oplog is large enough but the lag is still high, optimize performance by:

Running PLM closer to the destination to reduce network latency
Increasing CPU and memory on PLM host
Using faster hardware on the destination to improve write performance

Last update: June 5, 2025
Created: June 5, 2025