Hi, I'm hoping to get some help with a vexing issue with one of our keyspaces. During Reaper repair sessions, one keyspace will end up with hanging, non-started compactions. That is, the number of compactions as reported by nodetool compactionstats stays flat and there are no running compactions. Is there a way to determine which tables Cassandra is stuck on here?
Looking at graphs of pending compactions during Reaper sessions, the number of compactions will shoot up (as expected). The number will work its way down, and sometimes it will stop and plateau at a fixed level. A full compaction will get things going again, but we prefer to avoid those (even with the -s option). I realize there are various compaction tuning parameters regarding minimum age, tombstone percentage, etc but I need to know which sstables to look at first before blindly changing values. These are leveled compaction strategy tables FWIW. TIA! -richard
