When one of servers in my cluster goes down, volume alarm says some of "Actual Replication" fell in 1. Replication Factor in Volume Properties is 3, so Actual Replication shouldn't fall in 1.
Do someone hit on the reason?
My MapR version is v. 220.127.116.1155.GA on CentOS 5.7 (Final).
You are right, the replication should not drop to 1 when a single server goes down. However, if the volume was already underreplicated for some reason, then some containers may only have a single copy left. If you give the command : maprcli dump volumeinfo -volumename <volumename> -json, this will give a list of all the containers in the volume and the status of the replicas. Do this to find out which containers are having problems.
answered 23 Apr '12, 09:47