Hi Everyone We have migrated some of our clusters from Cassandra 3.11.11 to 4.0.1. We do repairs periodically triggered by some automation. Each time we run repair we do full `-full` sequential `-seq` primary `-pr` repairs for a portion of the full ring range and we finish iterating over the full range in a week.
With Cassandra 4.0.1 we started seeing snapshots created during repair accumulating until they eat up the disk space. We also see some errors with the message "Could not create snapshot at node x (put x's ip address). Check the logs on the repair participants for further details". Checking the logs of the mentioned node shows nothing. I have the following questions: 1. Did cassandra 3.11.11 perform automatic cleanup of snapshots? Why don't we see this on our Cassandra 3.11.11 clusters? 2. Is there a way to clear just repair snapshots? Repair snapshots are created with a GUID so it is difficult to use `nodetool clearsnapshot -tag` with a given tag. 3. If we do `nodetool clearsnapshot -all` while there is a repair job running, what will happen? Thanks -- Muhammad Soliman Senior Site Reliability Engineer [image: Booking.com] <https://www.booking.com/> Making it easier for everyone to experience the world.
