MongoDB 3.2.9: What happens if a chunk is dramatically larger than the 1GB max chunk size?

[a.fi...@celmatix.com]

I'm working on a MongoDB system which has chunks of ~1.2TB, with 6.6B documents, due to our choice of shard key, and I'm trying to figure out what the performance implications are. Has anyone worked on a cluster with chunks larger than the maximum allowed size?

[tom...@mongodb.com]

Hello,

I’m working on a MongoDB system which has chunks of ~1.2TB, with 6.6B documents

Could you post the output of sh.status(true)? The output would help to identify whether you have a chunk that is marked as ‘jumbo’. See also Tutorial: clean jumbo flag.
For MongoDB, the default chunk size is 64MB. You can increase or reduce this chunk size configuration. See Data partitioning with chunks.

due to our choice of shard key and I’m trying to figure out what the performance implications are

A chunk represents a single shard key value is a common cause for jumbo chunks. Since the chunk cannot be split, it continues to grow beyond the chunk configured size. The effect is that most of your read/write operations for a specified range will target a particular shard server; this may become a performance bottleneck.
Depending on your sharding use case, you may want to consider re-shard the data with a higher cardinality shard key for better load distribution. See also Considerations for selecting shard keys.

Regards,
Tom

​