How to configure which shards a sharded collection's data can reside?

Ops and Admin
,
#1

I have a sharded collection with 3 shards (shard1, shard2, shard3). Each shard is a 3-node replicaset (rs1, rs2, rs3).

I have a db called activity that has a large sharded collection called items. ie( activity.items). The data in this collection is split across shard{1,2,3}.

I have another db called app and collection called users (ie, app.users). This is not a sharded collection. It is housed on shard1.

I want to separate the replicasets that the apps.users and activity.items reside on. Ideally, app.users remains on shard1 and I can move activity.items from shard1 to a newly created shard4.

Is this possible? Any high level guidance on which commands to be looking at example docs would be greatly appreciated.

#2

I’m open to other solutions as well. For example, maybe there is a way to tag the shards or replicasets in such a way that there is no affinity for data to be migrated there?

#4

You define a zone which shard 2/3/4 belongs to, and then set up zone key range for activity.items.

Then the data on none-zone shards (e.g. shard1) will be gradually moved to other shards by balancer.

We are now using this zone support to isolate our services traffic.

#5

Thank you! And do you just define minkey/maxkey as the range for the zone?

#6

Yes, whole collection data should be in that zone, so full key range.

#7

Thank you, do you have an example of setting a full key range? Ideally for a compound shard key.