tag:blogger.com,1999:blog-2276603629708251265.post5219265926094105821..comments2023-09-18T13:32:20.547-04:00Comments on In Search of Updates: Ganglia Web UI Performance Improvement for Large Clustersbenhttp://www.blogger.com/profile/18412613940094299832noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-2276603629708251265.post-3927292932062225712015-01-21T05:11:36.478-05:002015-01-21T05:11:36.478-05:00Thank you very much, it is really helpful for us.Thank you very much, it is really helpful for us.Anonymoushttps://www.blogger.com/profile/17202319138576223945noreply@blogger.comtag:blogger.com,1999:blog-2276603629708251265.post-14700833489593219052014-05-13T21:54:22.025-04:002014-05-13T21:54:22.025-04:00Thanks for this cool article. And I have one quest...Thanks for this cool article. And I have one question:<br />>>edit /etc/ganglia-webfrontend/conf.php (or your local config file) and override the port used to talk to gmetad; specify 8661 instead of the default.<br /><br />In gweb2's conf, it use port 8652 by default, not 8651. So we should use 8662 instead of 8661 in your case?<br /><br />If I use 8661, ganglia web will not show summary value like CPUs total/Hosts up/Hosts down. <br />8662 seems OK.Nova Leohttps://www.blogger.com/profile/15837987462658312859noreply@blogger.comtag:blogger.com,1999:blog-2276603629708251265.post-82944845599939698262014-03-25T17:59:04.552-04:002014-03-25T17:59:04.552-04:00Thanks for the clarification.
I'been running ...Thanks for the clarification.<br /><br />I'been running gmetad 3.7.0 (pre-release) for a week, and yes, there are less gaps in my big clusters. But the slow front page loading is still an issue. In fact, that's the main reason I applied this improvement.<br />Unknownhttps://www.blogger.com/profile/04438800579008523901noreply@blogger.comtag:blogger.com,1999:blog-2276603629708251265.post-31041382937584555942014-03-25T17:51:02.205-04:002014-03-25T17:51:02.205-04:00Yes, it does duplicate the network traffic to the ...Yes, it does duplicate the network traffic to the gmond collectors. While this is unfortunate, network bandwidth (at least in my infrastructure) is not a bottleneck for ganglia data.<br /><br />You're correct that if you don't duplicate the data_source directives in each gmetad.conf file, it won't work. Both gmetad instances need all the data for this trick to work.<br /><br />I've heard rumors that the latest ganglia builds have a number of improvements that might make this approach unnecessary, though I haven't tested them myself.benhttps://www.blogger.com/profile/18412613940094299832noreply@blogger.comtag:blogger.com,1999:blog-2276603629708251265.post-9976312594834351882014-03-25T17:43:35.280-04:002014-03-25T17:43:35.280-04:00Thank you for the improvement. I've applied it...Thank you for the improvement. I've applied it and it works fine.<br /><br />My question is: Shall we duplicate the data_source directives in the new gmetad instance? Because I am noticing a duplication of network traffic in the gmetad box after the change. Aren't we polling each gmond twice with this improvement?<br /><br />I tried not to duplicate those directives, but the web UI home page doesn't show the clusters. May be the network traffic overhead is a necessary evil.<br />Unknownhttps://www.blogger.com/profile/04438800579008523901noreply@blogger.com