5 Replies Latest reply: Oct 17, 2012 4:14 AM by user12099976 RSS

    Cpu usage on a oracle cluster on aix 6.1 with no load.

    user12099976
      After installing a 4 node cluster 11.2.0.3 with 16 CPU's (4 on each node) on IBM 795 with aix 6.1 each server is using 0.5 cpu with no user load on the system.
      Running SIHA on one server typically uses 0.05 cpu with no user load on the system.

      Is this normal?

      Edited by: user12099976 on Aug 22, 2012 2:21 PM
        • 1. Re: Cpu usage on a oracle cluster on aix 6.1 with no load.
          Levi Pereira
          Hi,
          Who knows? Tell me that process is using this 0.5 of CPU.

          You can use topas or nmon (advanced) to find out who is using this cpu.

          Careful: You are using a LPAR so you must check how much resource is allocated to you use in this hosts. If you are looking to CPU on host you can see a wrong information.

          Check on HMC how much res you have allocated and available to this host.

          Regards,
          Levi Pereira
          • 2. Re: Cpu usage on a oracle cluster on aix 6.1 with no load.
            damorgan
            Your post just begs for the following questions.

            1. I doubt anyone has ever purchased a 795 with so few CPUs ... are you working in virtual machines hosted in LPars?
            2. When you use the term "CPU" are you referencing sockets, chips, cores, or threads?
            3. Where are you getting the numbers you posted for both cpu count and utilization? Name the specific tool(s).
            4. When you use the term "cluster" do you mean RAC or IBM clustering? ... you need to be specific.
            5. If RAC how did you implement storage? Using ASM? Are the ASM instances active? The ClusterWare nodeapps?
            6. No Oracle database instance is ever idle. At least not one that is open. Have you grepped active processes? What is active?

            After viewing your response I may be able to point you in the right direction but until then anything I or anyone else says is pure speculation.
            • 3. Re: Cpu usage on a oracle cluster on aix 6.1 with no load.
              user12099976
              1. Yes, there is a Oracle RAC cluster running on 4 LPARS (2 on site 1 and 2 on site 2). It's a stretched RAC cluster.
              2. Each LPAR is running with 4.00 cpus with smnt = 4 so nmon shows a logical CPU count of 16.
              Each LPAR have an entitlement of 1.00 cpus and when running the system with oracle clusterware, ASM, ACFS and one db instance with no user load nmon shows used cpu = 0.4 -> 0.5.
              3. I'm using nmon to get the number of cpus.
              4. I'm meaning a oracle RAC cluster
              5. I'm using ASM, it's streched so every thing is store at both sites. All 4 ASM instances are active. The ClusterWare are running only on the 4 DB nodes.
              6. The database is not idle, but I have no user load and were wondering how much overhead running RAC instead of SIHA is normal. It seems RAC uses 10 times as much CPU as SIHA running whith no userload.
              I have listed the process running with nmon, there is no process that uses all the cpu.


              When starting nmon it shows:

              TOPAS_NMON
              16 - CPUs currently
              16 - CPUs configured
              3720 - MHz CPU clock rate
              PowerPC_POWER7 - Processor
              64 bit - Hardware
              64 bit - Kernel
              10,q2lpar1004 - Logical Partition
              6.1.7.15 TL07 - AIX Kernel Version

              When showing long cpu nmon displays

              topas_nmon b=Black&White Host=q2dbs040 Refresh=2 secs 11:02.35
              Physical-CPU-Utilisation-Long-Term EntitledCPU= 1.00 UsedCPU= 0.407
              100%-||
              95%-||
              90%-||
              85%-||
              80%-||
              75%-||
              70%-||
              65%-||
              60%-||
              55%-||
              50%-||
              45%-||
              40%-||
              35%-||
              30%-|| s
              25%-|| s ss s ss s s ss s s sw s s s
              20%-|| s s s sss s s s s ssss s s s ss ssss s s s s sss s s s s s
              15%-|+ssssssUsssUUsssssssUsssssUssssssssssUsssUUssssUssUsssssUsssUsssssssssU
              10%-||UUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUU
              5%-||UUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUU
              0%------------------------------------------------------------------------


              nmon list of processes running.


              PID %CPU ResSize Char Command
              Used KB I/O
              6946826 1.7 102376 1 /grid/product/11.2.0.3/crs/bin/ocssd.bin
              6750414 1.6 26832 311 /grid/product/11.2.0.3/crs/bin/gipcd.bin
              8978560 1.4 63720 0 /grid/product/11.2.0.3/crs/bin/orarootagent.bin
              8323148 1.1 199024 0 /grid/product/11.2.0.3/crs/bin/crsd.bin reboot
              5439546 1.0 87124 0 /grid/product/11.2.0.3/crs/bin/ohasd.bin reboot
              7864544 1.0 45524 0 /grid/product/11.2.0.3/crs/bin/evmd.bin
              8912934 0.9 16820 0 /grid/product/11.2.0.3/crs/bin/octssd.bin reboot
              12976370 0.9 108412 0 ora_dia0_QSBE2
              9699486 0.5 80972 0 asm_dia0_+ASM2                                 
              29294728 0.5 94452 0 oracleQSBE2 (LOCAL=NO)
              8650874 0.5 135248 0 /grid/product/11.2.0.3/crs/bin/oraagent.bin
              10944700 0.5 100404 255 ora_lmon_QSBE2
              15073416 0.4 104356 0 ora_lms0_QSBE2
              2097216 0.4 960 0 gil
              6488140 0.4 85408 0 /grid/product/11.2.0.3/crs/bin/orarootagent.bin
              7340094 0.3 101420 1 /grid/product/11.2.0.3/crs/bin/oraagent.bin
              14352546 0.3 104384 0 ora_lms1_QSBE2
              9764952 0.2 72876 0 asm_lmon_+ASM2                                 
              7602270 0.2 20024 308 /grid/product/11.2.0.3/crs/bin/gpnpd.bin
              8192154 0.2 80508 0 asm_lms0_+ASM2                                 
              14090452 0.1 94468 0 oracleQSBE2 (LOCAL=NO)
              20250718 0.1 5804 0 /usr/bin/topas_nmon
              28835848 0.1 5304 0 /usr/bin/topas_nmon
              10158140 0.1 80708 0 asm_lmd0_+ASM2                                 
              15466502 0.1 104544 0 ora_lmd0_QSBE2
              6226076 0.1 163964 0 /grid/product/11.2.0.3/crs/bin/cssdagent
              18940090 0.0 88772 0 ora_vktm_QSBE2
              8061110 0.0 65392 0 asm_vktm_+ASM2

              Edited by: user12099976 on Aug 23, 2012 11:43 AM

              Edited by: user12099976 on Aug 23, 2012 2:10 PM

              Edited by: user12099976 on Aug 23, 2012 2:13 PM
              • 4. Re: Cpu usage on a oracle cluster on aix 6.1 with no load.
                user12099976
                There is not one single process that uses all the cpu, but the RAC processes are the top ones.

                The LPARS have enough CPU resources and if I put max user load on the system nmon reports used cpu = 4.0, as expected.
                • 5. Re: Cpu usage on a oracle cluster on aix 6.1 with no load.
                  user12099976
                  Added the following patch with PSU 11.2.0.3.4

                  13498267 - AIX-11202-CRS: HIGH CPU OVERHEAD OF AN "IDLE" RAC CLUSTER

                  Solved the problem