All I learned is that Cassandra will do resharding when a new node joins the cluster. Will Cassandra do resharding when a node gets too much data? If Cassandra has the resharding strategy, how to disable it?
2 Answers
Does Cassandra do resharding when a node stores too much data?
No, it does not. When a node stores too much data, the failure policies take over, and (usually) the node shuts down.
Adding or removing a node are the only times that partition token ranges are recalculated, forcing data to be reassigned to other nodes.
Another answer by @Aaron already explains that resharding only happens during adding or removing nodes. I want to add that you can use virtual nodes, so the partitions are smaller. However, you cannot change the number vnodes on already running physical node/machine.
To use vnodes change configuration in cassandra.yaml:
- Remove or comment out
initial_token - Uncomment and set
num_tokensto desired number of vnodes. 128 or 256 can be a good number.
Related