On Tue, Aug 27, 2019 at 11:28 AM Jacob Schlather <jschlat...@hubspot.com> wrote:
> I'm currently working on testing out the timeline consistent feature > > https://archive.cloudera.com/cdh5/cdh/5/hbase-0.98.6-cdh5.2.4/book/arch.timelineconsistent.reads.html > and I > had a couple of questions. > > 1. I setup region replication on a test cluster and everything seemed to be > working fine. I was doing get requests with timeline consistency and stale > results were coming back. Then something happened on the cluster and region > replication appears to be broken. When I look in the jmx endpoint on a > region server all of the replicas are empty and I added additionally > debugging information to the RegionServerCallable and the replica regions > are returning "The region's reads are disabled. Cannot serve the request". > I couldn't find anything errors that seemed to indicate what was wrong in > the region server logs. So, I was wondering what else I could do to debug > this and where I should be looking. > > This is interesting. This happens for any read against any region? Configs changed under the server and restart? Can you pinpoint when? Seems odd nothing in logs around this time. > 2. I've looked through the documentation and I might have missed something. > But I was wondering if there's any jmx metrics that measure the lag of the > region replicas or what would have to be done to add these metrics. Hbase > does expose replication metrics, but these appear to be for inter-cluster > replication rather than region replicas. > > I don't know of any read replica metrics. If you added any, they'd be welcome. Thank Jacob, S > Thanks, > Jacob >