We have a 32 node Cassandra cluster with approximately 150GB of data on each node. Running a full repair (not primary range) is taking approximately 26 hours to complete. This seems very slow. Any ideas on what I can check or tweak in the settings to speed up this repair.
Marshall
Besides repair with -pr option, the other idea is to use incremental repair if your Cassandra version is > 2.1. The first time incremental repair is slow as it needs to slit the SSTable into repaired and unprepared parts, but later on it would be faster as it won't repair what has been repaired. you can combine incremental/full repair, making incrental run more frequently and full repair less frequently.
