Sounds like the problem Ted Malaska was trying to solve with his multicluster client [1], though not sure that has gone anywhere for a while.
1. https://github.com/tmalaska/HBase.MCC On Monday, September 26, 2016, Sreeram <[email protected]> wrote: > Dear All, > > Please let me know your thoughts on viable approaches to fail over HBase > cluster across data centers in case of a primary data center outage. The > deployment scenario has zero data loss as one of the primary design goals. > Deployment scenario is Active-Passive. In case of active cluster being > down, there must be zero data loss fail over to the passive cluster. > > I understand that the built-in table level replication using 'add_peer' > might still lead to data loss since it is asynchronous. > > As a related note, is there is a way to specify the location (e.g. network > drive) where HBase WAL files in HDFS need to be written to ? The network > drive has synchronous replication across data centers. If the WAL files can > be written to the replicated network drives, can we recover in-flight data > in the passive cluster and resume operations from there ? > > Regards, > Sreeram > -- -Dima
