3 Replies Latest reply on Jul 13, 2018 4:30 AM by Simon Wardley

    BMC Discovery Excessive Swap usage and long reboot times - any recommendations?

    Simon Wardley

      We are using a disposable/temporary appliance for around 3 months making particular use of the start anywhere application modeling.

       

      Hosts: 997

      Network Devices: 74

      Applications: 93

       

      Appliance spec:

      Version: 11.2.02

      Ram: 15.6GB

      Swap: 30GB (but 24GB on 209GB attached drive in disk configurator)

      Processors Cores: 8 * Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz

       

      The appliance is very slow, we have only added the TKU's from march and a single custom pattern which is based around hostname standardisation and regex to define host environments etc.

       

      On a host restart the shutdown takes around 30 minutes to an hour to shutdown due to it turning off swap; which according to the performance graphs is 16GB in size!

      We are looking to attach an additional disk to segregate the swap to its own drive.

       

      Does anyone else have any recommendations to optimise ADDM?

        • 1. Re: BMC Discovery Excessive Swap usage and long reboot times - any recommendations?
          Andrew Waters

          Swap usage is not necessarily an issue - what is important is the amount of paging. The hardware performance graphs including information about paging. The swap usage graph on the same page tries to explain this:

          The appliance operating system attempts to make effective use of all available memory resources, both RAM and swap. To do this, it balances use of RAM between application data and system data used for caching and buffering. This stacked chart shows how the appliance's memory is partitioned.

          • The full height of the chart represents the total memory in the system, both RAM and swap. If the colored sections reach the top of the chart, all memory is full.
          • The lower part of the chart shows RAM usage, consisting of RAM used by application data, plus RAM used by the system, plus free RAM.
          • The top of the free RAM section indicates the boundary between RAM and swap. There is usually very little free RAM because the operating system uses anything spare for cache and buffer data.
          • The operating system often chooses to prioritize cache and buffer data over little-used application data, meaning some application data is placed in swap. Use of swap is therefore not a sign that there is insufficient RAM available.
          • If the amount of swap in use is often larger than the amount of RAM, system performance would probably benefit from having more RAM.
          2 of 2 people found this helpful
          • 2. Re: BMC Discovery Excessive Swap usage and long reboot times - any recommendations?
            Lisa Keeler

            I would suggest you should open a Support ticket.

            So many questions:

            What release version are you on?

            Does swap used keep increasing as time goes on? 

             

            On a support ticket, I would want these items initially on the ticket:

            1) screenshots of all the graphs found at Administration->Performance

            2) screenshot of the bottom of Administration->configuration page

            3) these logs from Administration->Appliance Support:

                     Performance

                     sar logs

                     system messages

            2 of 2 people found this helpful
            • 3. Re: BMC Discovery Excessive Swap usage and long reboot times - any recommendations?
              Simon Wardley

              Hi All,

               

              We had around 120 SAAM models and according to the latest appliance patch release notes.

               

              We have fixed some important bugs including excessive memory usage seen by customers with multiple SAAM models.

               

              Unfortunately the engagement has finished so we are not able to upgrade and see if the issue has been rectified by the patch.