Michael D. Posted April 15, 2012 Report Share Posted April 15, 2012 Hello, As you likely already know, we run CloudLinux which is a little different than standard Linux. There is a wealth of information on their site at http://www.cloudlinux.com/ if you're interested in learing more. CloudLinux, being different, calculates load a little differently than a standard Linux build. For example, we upgraded Cypress and Hermes from CloudLinux (2.6.18-338.19.1.el5.lve0.8.36) to CloudLinux (2.6.18-408.el5.lve0.8.58) tonight and the load numbers have changed drastically. The only thing that *really* changed was the way that CloudLinux calculates the server load. We just wanted to make our client base aware of this change as we do have a lot of customers who watch the load numbers as an indication of performance when in reality there are other much better and accurate resources to monitor. A post in the CloudLinux knowledgebase itself addresses the load changes which you can see at https://helpdesk.clo...rage-is-lowhigh Cypress was previously reporting loads between 5 and 9, which is 100% acceptable for a 16 core system and is now reporting anywhere from 15 to 20 which ordinarily would be very bad. In this instance, it'a mis-calculation of the load and has no bearing on the real performance of the server. If you have any questions about this, by all means, post a response on this thread and I'll be happy to address it. Quote Link to comment Share on other sites More sharing options...
Michael D. Posted April 16, 2012 Author Report Share Posted April 16, 2012 I got with CloudLinux yesterday and they told me that this is **definitely** unexpected and not intended. They said the part of the kernel that calculates the load averages is modular and can be changed out without a reboot, and that they anticipate having it nailed down hopefully by Tomorrow (Tuesday). Quote Link to comment Share on other sites More sharing options...
Vilandra Posted April 16, 2012 Report Share Posted April 16, 2012 Thank you Mike! Much appreciated! Quote Link to comment Share on other sites More sharing options...
Michael D. Posted April 17, 2012 Author Report Share Posted April 17, 2012 Keep in mind the change is cosmetic, the boot into the new kernel didn't cause performance to go down nor will changing the load module cause performance to go up. There were the same amount of accounts on the server before and after the reboot . Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.