Home/ Questions/Q 7576925
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-05-30T17:03:17+00:00

For my needs, I have to do pre-splitting but my shard key is a

  • 0

For my needs, I have to do pre-splitting but my shard key is a compound key. both keys of my compound key (k1 and k2) are int and my desired chunk structure is something like this

chunk1 : k1 = minkey to 1, k2 = minkey to maxkey

chunk2 : k1 = 1 to 2, k2 = minkey to maxkey

chunk3 : k1 = 2 to 3, k2 = minkey to maxkey

chunk4 : k1 = 3 to 4, k2 = minkey to maxkey
…

So what I really want to do is split on the first key only and leave the second key to include all possible ranges but if run this 

db.runCommand( { split : "db.mycollection" , middle : { k1: 1} } );

it will fail saying the full shard key is not provided so it makes me to call this

db.runCommand( { split : "db.mycollection" , middle : { k1: 1, k2:somenumber } } );

this command works but the resulted split is bad because it gives me chunks like below

chunk1 : k1 = minkey to 1, k2 = minkey to somenumber

chunk2 : k1 = 1 to maxkey , k2 = somenumber to maxkey

this is bad because if users insert something a document where k1<1 and k2>somenumber, there is no chunk in the system that covers it and the insert should fail.

How can I create such chunk boundaries with my compound key?

NOTE: One may ask why I have k2 if it is always minkey to maxkey. I have k2 like that to allow further splitting on k2 for same k1 in future if needed be ( and I am sure it will be needed on some of chunks)

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    This should do it :

    db.runCommand( { split : "db.mycollection" , middle : { k1: 1, k2:MaxKey } } );
db.runCommand( { split : "db.mycollection" , middle : { k1: 2, k2:MaxKey } } );
db.runCommand( { split : "db.mycollection" , middle : { k1: 3, k2:MaxKey } } );
db.runCommand( { split : "db.mycollection" , middle : { k1: 4, k2:MaxKey } } );

    By the way, you can never create a situation where a certain range is not covered by chunks since all you’re doing is splitting existing ranges. Also note that k2 cannot be from MinKey to MaxKey technically. Your range for the first chunk is going to be {k1: MinKey, k2: MinKey} to {k1:1, k2: MaxKey}. The next one is going to be {k1: 1, k2: MaxKey} to {k1:2, k2: MaxKey} and so on

    • 0