hostserver CPU usage

Clovertech Forums Read Only Archives Cloverleaf Cloverleaf hostserver CPU usage

  • Creator
    Topic
  • #48426
    Kevin Scantlan
    Participant

      It seems that our hostserver is using too much CPU.  We jump up to 100% on the machine, so we drop the hostserver and this relieved the CPU crunch quite a bit.  After a while we turn it back on until the CPU usage climbs and we have to do the same thing again.  Here’s an example of a top command:

      CPUs:  4; load averages:  4.19,  3.53,  3.34    09:28:30

      94 processes:  83 sleeping, 8 runnable, 3 zombie

      Memory: 4516M Total. Real: 2052M, 54M Free, 1261M Buffers. Virtual: 2464M, 2250M

      Free

       PID USERNAME PRI NICE   SIZE   RES STATE   TIME   WCPU    CPU COMMAND

      16274 hci       98   4  651k  172k  run  8+09:34  1.7/ 5.4 lm_ip

      21970 hci       70   4   17M 9500k  slp 63:40:20  0.5/ 1.7 hciengine

       4992 root      60   0  161k   36k  slp 14:57:55  0.1/ 0.4 syncd

      19148 hci       69   4   23M   15M  slp 13:27:32  0.1/14.2 hciengine

      15320 hci       68   4  141M  141M  slp 11:53:20  0.1/50.0 java

       7018 root      60   0   12M 8372k  slp  2:53:41  0.0/ 0.1 dsmc

      22544 hci       93   4   41M   39M  slp  2:21:53  0.0/10.1 hciengine

      32422 hci       68   4   21M   20M Fslp  2:10:01  0.0/ 2.3 hciengine

      There are 2 percentages associated with the java command: 0.1 and 50.0.  Does anyone know how to interpret those values in the WCPU column?

      Thanks.

    Viewing 11 reply threads
    • Author
      Replies
      • #58604
        Pete Gilbert
        Participant

          We have encountered a similar situation here. AIX 5.2, Qdx 5.3r2. On one of our engines the host server runs amok several times a week. Usually, it will stop after a ten minute spasm, but yesterday it ran full tilt for a couple of hours before I bounced it. The short spasms are just long enough to trigger our CPU alerts, and our operators have started to ignore them, becuase it clears itself up after a few minutes.

          Is there anything that I should be looking at as a potential cause for this behaviour?

        • #58605
          Anonymous
          Participant

            Check your server ini file, and what is the jvm args set for?

            it should be at least 256M.

            Thanks

            reggie

          • #58606
            Pete Gilbert
            Participant

              I got this from another post. The jvm arg is set as:

              jvm_args=-Xmx256m

              Yesterday, I counted five spasms, lasting from 8 to 14 minutes.  We’ve had two so far, today.

            • #58607
              Eileen Pitts
              Participant

                I am on QDX5.3 running on AIX5.2 and I am having the same problem.  I looked at $HCIROOT/server/server.ini and I don’t have an entry for jvm_args.  Should I have this and where should it sit in this ini file?

              • #58608
                Mark Gathers
                Participant

                  Hello all,

                  QDX 5.3 rev3, AIX 5.2

                  Regarding the java arg,

                  Set as follows in the server.ini file

                  [general]

                  jvm_args=-Xmx256m

                  [logging]

                  Check all workstations accessing the engine and repeat same procedure  to the client.ini file.  Also, include the netmon parm.

                  [general]

                  doc_base_dir=C:quovadxqdx5.3integrator

                  debug=false

                  jvm_args=-Xmx256m

                  netmon_auto_load_logs=false

                • #58609
                  Pete Gilbert
                  Participant

                    I sent the client ini change out to all of the developers here. It seems to have helped, but we are still having one or two spasms per day. These spasms only occur between 0900 and 1700 on weekdays, so I’m pretty sure it is something that one of our developers is doing that is causing the spasm. Curiously, we have five other interface engines and have not experienced these sorts of spasms on them.

                    Can anyone think of anything else that I should look at?

                  • #58610
                    Anonymous
                    Participant

                      Pete,

                      if you had made the change then I would bounce the hostserver and the site. Start them cleanly.

                      Another thing I have noticed that, does anyone using shortcut to hciaccess or hcinetmonitor

                      In windows and C:quovadxqdx5.3integratorclguibinunsupported

                      If anyone using the shortcut to hcinetmonitor or anything else, the panic occurs too.

                      Therefore it is advisable to use the IDE to access the Netmonitor or Netconfig, testing tools.

                      Hope this helps.

                      thanks

                      Reggie

                    • #58611
                      Rick Brown
                      Participant

                        What Mark says…

                        The most important being this line:  netmon_auto_load_logs=false

                        Make sure that gets put in all the client.ini’s on each client machine that connects to the host server.

                        Thanks,

                        Rick

                      • #58612
                        David Harrison
                        Participant

                          Cloverleaf 5.2 on Solaris 8

                          Is this true for Cloverleaf 5.2? My server.ini file has no entries under [general] and my client.ini file has the line

                          Code:

                          jvm_args=-Dsun.java2d.noddraw=true

                        • #58613
                          Rick Brown
                          Participant

                            Yes David – those settings will work for 5.2 as well.

                          • #58614
                            Greg Eriksen
                            Participant

                              Hi Rick,

                              Just wanted to confirm that in 5.4 the netmon entry is no longer necessary.

                              The following is from the CIS54relnotes:

                              3.53

                              netmon_auto_load_logs=false in client.ini should be set at install

                              Previously, if netmon_auto_load_logs did not exist in client.ini, then the IDE considered it as

                            • #58615
                              Rick Brown
                              Participant

                                Yes, Greg is correct.  5.4 or later will be set to off by default when the client is installed.

                            Viewing 11 reply threads
                            • The forum ‘Cloverleaf’ is closed to new topics and replies.