Hi Nick, I don't know, what would be better/easier? I'd think per RS may be OK, but you'll know better about where HLogs can be "caught".
Btw. that Databus project just happened to spark the idea, but I guess the general version of what I'm thinking is simply: "Can we introduce some HLog Listener interfaces/hooks so people can plug in there and do whatever they want with HLog entries". One such impl. can be for putting HLog entries on that Databus. Otis -- HBASE Performance Monitoring http://sematext.com/spm/hbase-performance-monitoring/ On Mon, Mar 4, 2013 at 2:39 PM, Nick Dimiduk <[email protected]> wrote: > Hi Otis, > > This is a cool idea. Would you see a single event pipe containing all > edits, per RS pipes, something else entirely? > > -n > > On Sat, Mar 2, 2013 at 7:08 PM, Otis Gospodnetic < > [email protected] > > wrote: > > > Hi, > > > > I was looking at > > > > > http://engineering.linkedin.com/data-replication/open-sourcing-databus-linkedins-low-latency-change-data-capture-systemand > > then at > > > > > https://github.com/apache/hbase/tree/trunk/hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserverand > > wondering... > > > > .... would it be possible/interesting/better to "open up" a little bit > and > > implement something that puts HLog events on a bus like that Databus and > > thus allow *anyone* to do *anything* with data being written to (or > deleted > > from) HBase? Replication itself, secondary indexing, etc. > > > > Thanks, > > Otis > > -- > > HBASE Performance Monitoring > > http://sematext.com/spm/hbase-performance-monitoring/ > > >
