Chet - can I put this in the Community FAQ? -- James Pulver Information Technology Area Supervisor LEPP Computer Group Cornell University
Chet Luther wrote, On 4/18/2009 8:20 PM: > On Apr 17, 2009, at 4:12 PM, mwcotton wrote: >> I see the ability to have multiple zenhub workers, whats the school >> of thought on when to add a worker? Is there a log I can look at? Is >> it a cpu or memory issue? Seems like the more the better, but in the >> computer world that doesnt always hold true. > > The default configuration is to not use workers at all. What you > should be on the lookout for in this configuration for that would tip > you off to enable workers would be regular "Timeout connecting to > zenhub: is it running?" messages when you attempt to remodel devices > or in other areas. > > So now.. now many workers? There's no right answer for every possible > scenario, but there are some good rules of thumb. The first would be > that you never want to exceed the total number of CPU cores minus one. > So on a 4 core box, you wouldn't want more than 3 workers. This is to > keep the web interface responsive. > > Maybe more importantly, you could start by enabling 2 workers. Then > you can keep an eye on the size of your zenhub queue. This isn't one > of the standard collector metrics right now, but you can enabled it by > following the following steps. > > 1. Go to http://yourzenoss:8080/zport/dmd/Monitors/perfConfig > 2. Click into the PerformanceConf template > 3. Add a new data source. The name is zenhub and the type is Built-In > 4. Add a data point to this zenhub data source called workListLength > 5. Add a new graph to the PerformanceConf template > 6. Add the zenhub_workListLength to this graph. > > Now when you click on the Performance tab of your collector you will > be able to see what the size of the zenhub queue has been over time. > You should never see it go over 50, and it should stay below 10 the > vast majority of the time. If this is not the case, you can increase > the number of zenhub workers up to the total cores minus one ceiling > mentioned above. > > If you get to this point and your zenhub queue is still too high, your > server likely needs more horsepower. > _______________________________________________ > zenoss-users mailing list > [email protected] > http://lists.zenoss.org/mailman/listinfo/zenoss-users _______________________________________________ zenoss-users mailing list [email protected] http://lists.zenoss.org/mailman/listinfo/zenoss-users
