Alexander Dejanovski

We do have a `millisSinceLastRepair` per repair unit (== schedules) since a bit of time now. Getting a metrics on the number of times it completed would be a little...

The ordering of the repair by cluster table was wrong in 2.1 and has been modified since then. It's highly probable that this issue doesn't exist anymore in later versions....

Since 2.2.0, Reaper will allow concurrent segments from different repair runs to be running on the same nodes. To avoid overwhelming nodes, a new setting was introduced named `maxParallelRepairs` which...

Now that I think of it, there could be some race conditions in `EACH` mode since there's no consensus on the parallel repairs limiter. One Reaper instance could have started...

@afsanchez83, we've removed the postgres backend since 3.0 which will no longer get fixes or improvements. Would it be possible for you to switch to the Cassandra backend and assess...

Reaper's default [changed a while ago](https://github.com/thelastpickle/cassandra-reaper/commit/3c46245385db54c5f0567080014a27f11e08ddf7) as we were seeing issues with 16 segments per node, which tends to generate timeouts. 64 was considered a more conservative value over 16....

ouch... Could you share the content of the `running_repairs` table after obfuscating the IPs (but keeping them consistent so we know which node is which). Are there segments in running...

Error messages could be related to the truncation as we removed the existing locks. Still blocked repairs like you've seen should never happen and need to be thoroughly investigated.

Hi folks, is this still an issue you're facing?

@LarsinYousif, the postgres backend has been removed in Reaper 3.x and it is highly unlikely that we'll come up with a fix for it. Is there any chance you could...