3 Replies Latest reply on May 31, 2016 7:01 AM by Russell Christopher

    tabmon workbook

    Jeff Strauss

      Have you developed any really great vizzes against tabmon metrics?  If so, can you share? 

       

      We are having a look at the one included as part of tabmon and it doesn't seem to really tell a clear story...  we're only interested in cache effectiveness, averagerequestlatency, requestsfailed, requestsprocessed and heapmemoryusage.

       

       

       

       

      thank you, Jeff

        • 1. Re: tabmon workbook
          Russell Christopher

          Heya Jeff -

           

          Some of these metrics aren't really meaningful (or in some cases, actionable). For example, the cache JMX Mbeans are fun to look at, but there's really nothing you can do change the behavior of the server to make the numbers in the viz themselves change. I typically ignore this dashboard.

           

          As I recall, AverageRequestLatency focuses on how long it takes to fulfil "internal" requests to get in and out of the postgres database. It doesn't really have much to do with render time of "real" vizzes. I guess this metric could be useful to know if the disk was bad or the postgres database was getting overwhelmed, but that's about all I can think of. RequestsProcessed/RequestsFailed are pretty cool, but the actual "tasks" they monitor have changed in 9.3, so I don't know if you want to spend too much time with 'em

          2 of 2 people found this helpful
          • 2. Re: tabmon workbook
            Jeff Strauss

            thanks Chris!  Here's what we ended up developing in terms of latency and cache.

             

            While we agree that they're not so that is actionable, it is interesting to expose the metrics to see the effect on the overall platform.  For example, within our caching tab, it's interesting to see that there is zero activity flat line within "native external query cache" and that image cache hits are flat line within one of our workers (ord-tbl104) .  And the fact that there are not so many image hits tells us that storing image tiles within RAM may not necessarily be worth it as there aren't so many repeat lookups for the same viz.

             

               

             

            1 of 1 people found this helpful
            • 3. Re: tabmon workbook
              Russell Christopher

              A relatively low cache hit ratio on the Native Query Cache is pretty normal since we check the Abstract Query Cache first. If you do / did more server-side rendering AND you have hard-coded dashboard sizes, you'd / you'll probably see more tile cache hits.

              2 of 2 people found this helpful