Hi Mathias, Keep in mind that they are kept as MetricsTimeVaryingRate meaning they trail by "period" seconds. The values are accrued over the poll interval and then updated at the end.
But you are saying there is a huge difference? I think this is due to the puts not actually using puts anymore, but the "multi" calls, ie. multiPut and multi. Could you verify the numbers you see there? Lars On Wed, May 18, 2011 at 4:48 PM, Mathias Herberts <[email protected]> wrote: > While on the topic of metrics, some of them seem misleading, > > in my setup, 'hbase.metrics.put_num_ops' does not seem to reflect the > number of put ops received by a RS, it's simply way below and seems to > only reflect puts done in .META., is that supposed to be the case? > > > On Wed, May 18, 2011 at 16:42, Stack <[email protected]> wrote: >> MIA I'd say. They look useful. >> St.Ack >> >> On Wed, May 18, 2011 at 6:34 AM, Lars George <[email protected]> wrote: >>> Hi, >>> >>> Did anyone else notice that the RPCMetrics are supposed to export four >>> internal values: >>> >>> public MetricsTimeVaryingRate rpcQueueTime = >>> new MetricsTimeVaryingRate("RpcQueueTime", registry); >>> public MetricsTimeVaryingRate rpcProcessingTime = >>> new MetricsTimeVaryingRate("RpcProcessingTime", registry); >>> public MetricsIntValue numOpenConnections = >>> new MetricsIntValue("NumOpenConnections", registry); >>> public MetricsIntValue callQueueLen = >>> new MetricsIntValue("callQueueLen", registry); >>> >>> The first two show just fine on JMX and Context based output, but the >>> latter two don't. Nowhere. >>> >>> Anyone having a clue where they went MIA? >>> >>> Lars >>> >> >
