On Mon, Dec 14, 2009 at 2:18 PM, Lars George <[email protected]> wrote: > Stack > > +1 > > Thought about the same. > > On Mon, Dec 14, 2009 at 7:24 PM, stack <[email protected]> wrote: > >> Edward: That looks sweet. Should we add a pointer or text of your howto to >> our metrics doc page (See >> http://hadoop.apache.org/hbase/docs/r0.20.2/metrics.html)? >> >> Thanks, >> St.Ack >> >> On Sat, Dec 12, 2009 at 7:10 PM, Edward Capriolo <[email protected] >> >wrote: >> >> > All, >> > >> > I have created cacti graphs for all the JMX exported variables with >> HBase. >> > >> > HBase 20.2 - RegionServer - Atomic Incr >> > HBase 20.2 - RegionServer - BlockCache >> > HBase 20.2 - RegionServer - BlockCache Count >> > HBase 20.2 - RegionServer - BlockCache HitRatio >> > HBase 20.2 - RegionServer - Memstore Size >> > HBase 20.2 - RegionServer - Regions >> > HBase 20.2 - RegionServer - Storefile Index Size >> > HBase 20.2 - RegionServer - StoreFiles >> > HBase 20.2 - RegionServer - Stores >> > HBase 20.2- HMaster - MasterStatistics >> > >> > You can check them out here: >> > http://www.jointhegrid.com/hadoop/ >> > >> >
You can definitely add any information. I am even willing to just have the code incorporated upstream / move to hbase as (contrib) . The Challenge: In hadoop-core someone did contribute a command line NameNode "fetch tool". My code does not build against hadoop/hbase, but works exactly like the fetch tool. The majority of the work is on the cacti side in designing intelligent template displays for the data. For example, the "Atomic Increment" graph displays four items into one graph, rather then displaying four separate graphs. Much of the work is not unit testable and if someone introduces a new variable new graphs are not created manually. For now I will add some information to the page you suggested telling users if they have turned on JMX the information can be gathered with this tool. If you want to discuss ways to get it upstream, that is great as well. -Ed
