Hi,
Yes, it needs to run on every node. This may help -
https://www.datastax.com/dev/blog/repair-in-cassandra
Consider using '-full' to avoid incremental repair whatsoever.
Check the -dc, -dcpar and -local options for controlling its behaviour
across DCs please -
https://docs.datastax.com/en/cassan
Taking a step back and in case you are hosted on a cloud - We are entirely
on AWS and upload snapshots, and incremental files periodically, to S3 as
part of the backup job. It gives us durability and the freedom to clean up
from locale sooner than otherwise would require. Currently, we also are
try