One data table can be split into multiple tablets .
Why can tablets be split into Parent tablets and Child tablets?
What is the relationship between parent tablets and child tablets?
Is there a specific data instance to show?
Thanks a lot.
Hi, There’s nothing like parent and child tablets.
The SQL tables are sharded on their primary key (and indexes on their key) to distribute the rows to tablets. Those are distributed and replicated on the cluster.
Hi， thank you for your answer.
But, I mainly don’t understand the logical relationship between parent tablet and child tablet. Can you provide a specific scenario and examples to help me understand?
Thanks a lot.
Please link to the page where you read about “parent/child tablet” so we know the context.
I am reading the source file tablet_split_manager.c.
I saw some descriptions of parent and child tabs in the relevant functions.
The context here is on “automatic tablet splitting” https://docs.yugabyte.com/preview/architecture/docdb-sharding/tablet-splitting/#automatic-tablet-splitting.
Once a split has been executed on a tablet, the two resulting tablets require a full compaction in order to remove unnecessary data from each. These post-split compactions can significantly increase CPU and disk usage and thus impact performance. To limit the impact, the number of simultaneous tablet splits allowed is conservative by default.
So the “parent” tablet is the tablet before splitting, and the “child” tablets are the 2 created ones. This “parent->child” relationship exists while the “splitting” is happening. When the splitting process finishes, the “parent” tablet no longer exists and the child tablets are just normal tablets.
I understand now.
Thank you very much.