Dangers of ((hcimsiutil -zt)) on running thread??

Clovertech Forums Read Only Archives Cloverleaf Cloverleaf Dangers of ((hcimsiutil -zt)) on running thread??

  • Creator
    Topic
  • #51326
    Troy Morton
    Participant

      Does anyone know of any bad things that could happen if you run..

      Code:

      hcimsiutil -zt


      … on an actively running thread/process/site?

      I tried it in my development test site and it appears that the threads go into an “Initializing” state until they receive the next interface message … that’s all.

      I’m writing a script to colllect thread statistics.

    Viewing 7 reply threads
    • Author
      Replies
      • #69725
        Jim Kosloskey
        Participant

          No problems.

          The ‘initializing’ Status is temporary. A lot of activity will cause it to revert to the actual connection status.

          There are two schools regarding clearing stats after collection.

          If you do clear after each collection then you have no continuum.

          If you collect periodically throughout the day and then clear only once per day you have a daily continuum – but – the analysis process (that thing receiving the stats) needs to be able to do delta reporting.

          Having had a past life in the perfomance/capacity analysis/planning arena, the generally accepted data collection method was to clear on a periodic basis and collect in between (the continuum concept). You will note that is the way msi works – it just keeps collecting until it is cleared. This allows for a more usable set of metrics for performance, capacity, etc. – however it requires heftier capability of the reporting medium.

          I have seen some success for management reporting doing the clear after collection method, but it presents some performance/capacity analysis issues.

          So it really depends on your intent.

          If your intent is to support both (that is often a stated goal) then the continuum philosophy has more merit.

          An additional note if your intent is to do performance/capacity analysis/planning – it is important you establish and periodically evaluate your peak demand period.

          That is different for different workloads (batch versus real-time for example) in that the time period for a peak will differ between batch and real-time. Real-time for example might look at no more than 15 minute periods to determine peak demand. Batch on the other hand might look at periods of one hour or more.

          Also understand a peak demand period is NOT the peak usage period. They frequently do not even coincide.

          Rather the peak demand period is based on arrival rates.

          Once the peak demand period is established (ideally for no more than one or two periods on one day of the week), simply monitoring that period can provide the information needed to track performance/capacity and provide one basis for capacity/performance forecasting. This then reduces the effort ot analyze perfromance/capacity.

          There is a lot more to this. Usually folks don’t realize the scope until after they begin collecting the data – but this efffort is like Pandora’s box – once opened it seems to keep growing.

          Good luck!

          email: jim.kosloskey@jim-kosloskey.com 29+ years Cloverleaf, 59 years IT - old fart.

        • #69726
          Troy Morton
          Participant

            At this time I’m just collecting for Management reporting.

            Also (to a lesser degree) for long-term server capacity planning.

          • #69727
            Russ Ross
            Participant

              We have been zeroing out thread counts via

              hcimsiutil -zt $thread

              near midnight from the very beginning.

              After being woke up repeatedly with false alerts I eventually figured out the initializing status on the thread was triggering the false alerts.

              We solved our problem by cycling the process log file via

              hcicmd -p $proc -c ‘. output_cycle’ > /dev/null 2>&1

              shortly after zeroing out the thread counts which causes the initializing status to disappear and eliminates the false alerts.

              Russ Ross
              RussRoss318@gmail.com

            • #69728
              Troy Morton
              Participant

                Awesome.  Thanks Russ.  I’ll definitely keep this in mind.

              • #69729
                Troy Morton
                Participant

                  Okay, I have a script in place that is getting message/byte/queue/error/activity information from threads.

                  However, we have another cron that cycles SMAT and log files, then archives them.

                  So, I don’t want my script to cycle the process logs.  Anyone know of another command/activity that I can issue to the thread to get the “initializing” status to go away?

                  Thanks,

                  Troy

                • #69730
                  James Cobane
                  Participant

                    Troy,

                    You can simply execute a ‘start’ on the thread to refresh the status.

                  • #69731
                    Troy Morton
                    Participant

                      That worked!  Thanks.  I probably should have been able to think of that myself.

                    • #69732
                      Keith McLeod
                      Participant

                        hcimsiutil -X or hcimsiutil -xt can now be used in versions at least 5.7 not sure of 5.6.  This zeros out counters but leaves date and time stamps.

                    Viewing 7 reply threads
                    • The forum ‘Cloverleaf’ is closed to new topics and replies.