configuration error. NullContext for rpc metrics was still used instead of NullContextWithUpdateThread and I kept failing retracing my steps because min and max metrics were working even then, strangely enough.
I am still having 0 as average times but I am prepared to believe it indeed might be under 1ms in this particular case. But at least now i see per-method non-zero rpc counters. On a completely other issue i was reporting, I still have no ideas why remote client latencies would hover around 25ms in this case if they avg <1ms on the server. But i am inclined to believe it is unlikely to be a region server problem, as evidenced by rpc metrics. At least it doesn't happen with region server on my local machine loaded with exactly the same data and running the same test locally. It seems to be more or less isolated to a network issue of some sort. I will stage a number of cleaner tests to reduce network IO interference from other activities on that subnet to get a better idea. This cluster is not only a newer hbase version, it is also a new hw setup compared to what we know was working for us. Too many factors. On Wed, Apr 20, 2011 at 11:56 PM, Stack <[email protected]> wrote: > What was the solution? > St.Ack > > On Wed, Apr 20, 2011 at 11:15 PM, Dmitriy Lyubimov <[email protected]> > wrote: >> ok, i think i solved the mistery. thanks for help. >> >
