9 Replies Latest reply on Mar 28, 2011 10:35 PM by KarimK

    Profiler performance - CPUs hit 100%

    j_butterfield

      Has anyone seen poor performance of Profiler when run on a single server?  I noticed some previous posts were removed about this subject and was hoping to get some more information.  We were monitoring 4 netapps, one VC with 13 ESX (87 VMs), and three fiber switches and the CPU kept pegging at 100%.  This is one a quad core box with internal disk and 8GB RAM.  I don't want to add additional physical boxes to my environment or kill a esx server by pegging the CPU on it.

      Thanks,

      John

        • Re: Profiler performance - CPUs hit 100%
          lchance

          We are getting 100% CPU with our evaluation install and we just made a big purchase of Profiler this week.

          I need to get a handle on this fast before I convert it into production.

            • Re: Profiler performance - CPUs hit 100%
              j_butterfield

              I was told by support that the best configuration is to use a second box for the collector service.  That will drop the CPU usage on the reporting box.  Does require a second/third/forth box especially if you are polling Netapp devices.

                • Re: Profiler performance - CPUs hit 100%
                  KarimK

                  Is there any documentation / info on how to setup the collector service to run on a different host than the main core Profiler server?  I have the same issue with the collector service hammering the CPU regularly and making the application non responsive.

                    • Re: Profiler performance - CPUs hit 100%
                      bmrad

                      Yes, you can download agents from the website (if you are evaluating) or your portal (if you have purchased).  Agents are installed on OS (window, linux, unix) and then can be leverage to poll other devices and applications.  If you look on page 18 of the installation guide, it will tell you how to install the agents.

                      http://www.solarwinds.com/support/profiler/docs/Profiler_Installation_Guide.pdf

                      Brian

                        • Re: Profiler performance - CPUs hit 100%
                          KarimK

                          Yes, I'm already doing that - the Profiler server itself is not the poller for most of the devices - that function is spread across many different agents.

                          What I'm asking about is that the previous post implied that you could run the actual Profiler Collection module (ProfilerCollector.exe) on a separate server than the rest of the Profiler server services (the MySQL database, the WebServer, the Maintenance module etc) since it's the Collector service that constantly pegs the CPU on the server at 100%, which makes the application painfully slow to use.

                          Is that actually possible and if so how do you configure that?

                           

                          Karim

                            • Re: Profiler performance - CPUs hit 100%
                              bmrad

                              Karim,

                              What is the Profiler Server hardware (#CPU, GB of Mem)?  How many arrays, devices and VM are you collecting from?  What OS (Windows 32bit, Windows 64bit or Linux)?

                              Are you running data retention and cleanup script on a regular basis?

                              Generally, high collector CPU means underpowered hardware or a corrupted table... it might be worth the support guys to take a look at this.

                              Brian

                                • Re: Profiler performance - CPUs hit 100%
                                  KarimK

                                  Hi,

                                  The Profiler server is a Virtual.   Win 2008R2.  3 Virtual CPU, 16GB RAM.  It only using about half the RAM so that seems ok.  It  previously had two vCPU's allocated, the third didn't seem to make any difference.

                                  Underlying ESX 4.0 host is a Dell R710, Dual Xeon X5570 2.93Ghz, 96GB RAM.

                                  Profiler tells me that ESX server is running about 46% free processor time.

                                  Have agents collecting for:  14 NetApp, 2 x HDS, 8 x LSI, 1 x IBM SVC, 3 x Cisco MDS & currently polling 31 ESX servers running about 450 guests.

                                  Haven't run the maintenance script lately - didn't seem to make any difference last time.

                                  I've logged a call previously regarding high CPU usage - 202566.

                                  Server and agents are all running 4.12.4

                                  Karim

                          • Re: Profiler performance - CPUs hit 100%
                            bmrad

                            We you able to adjust your configuration to improve performance?

                            Brian

                        • Re: Profiler performance - CPUs hit 100%
                          Garrett Gross

                          If the issue is not resolved by upgrading the agent, please submit a support ticket at www.solarwinds.com/support/ticket/ and a member of our support staff should be able to help you out.