Clovertech

Swap Space Consumption Increase CIS5.6 Rev2 on AIX5.2

› Clovertech Forums › Read Only Archives › Cloverleaf › Cloverleaf › Swap Space Consumption Increase CIS5.6 Rev2 on AIX5.2

  • Creator
    Topic
  • July 28, 2008 at 6:36 pm #50211
    Bob Richardson
    Participant

      Greetings,

      Folks, we have just upgraded to CIS5.6 Rev2 on AIX5.2 maintenance level 8 and are experiencing an incremental increase in the use of swap space. The rate is 1% per 24 hour period.  We have multiple sites doing about 81 million messages a month (inbound and outbound combined).

      Per our Operating System Group no paging is taking place even though utilization is going up.  We suspect a memory leak – this has been reported to Healthvision tech support.

      We have tried cycling the hostserver; monitor daemons for each site, and any identified *large* processes with no effect.  We have https with SSL connections too and cycled those as well – no effect.  Our ulimit settings are at the recommended levels.  Our hardware is an IBM P630 with 4 CPUs with a clock speed of 1454 MHz.  We are running in 32 bit mode.

      Has anyone out there experienced this swap space utilization problem?

      Your solution?  Recommendations?

      Thanks for any help – if we determine cause with Heathvision it will be posted for your reference.

    • Creator
      Topic
    Viewing 0 reply threads
    • Author
      Replies
      • April 22, 2009 at 6:50 am #65225
        Richard Hart
        Participant

          Hi Bob.

          We have a few sites on 5.6 rev 2 and AIX 5.3 (the rest are on 5.4.1).

          The sites on 5.6 are the using web service  (with IB 2.1) and I think I have found a memory leak that has been causing some issues for us.

          Some of our web service threads send in excess of 200,000 records per day and we have frequent crashes.  These are batch loads, Operations restart the site and nothing is lost, so the business has not been pushing a fix.

          We use IbEnvelopeIn, IbServerEnvelopeOut and IbClientEnvelopeOut  and these routines do not have a ‘ibmimedestroy’ call after an ibmimecreate.

          I’ve modified the file (ib.tcl) and the memory leak is considerably reduced, however it is still there.

          I have raised this with support via our local rep.

      • Author
        Replies
      Viewing 0 reply threads
      • The forum ‘Cloverleaf’ is closed to new topics and replies.