Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
Description
ecn suggested to me that the replication servlet could be updated to be more clear about what is healthy and what is not. Update the servlet to compute a general measure of "healthy", "potentially unhealthy", and "not healthy at all".
Generally speaking, we have to temper the amount of replication backlog in with the size of the system we're working against. A backlog of 50 files could be expected when running with 25+ tservers, while 15 files would be indicative of a problem with only 2 tservers.