AIX 5.3/Cloverleaf 5.5rev1 Connectivity Issue

Homepage Clovertech Forums Read Only Archives Cloverleaf Cloverleaf AIX 5.3/Cloverleaf 5.5rev1 Connectivity Issue

  • Creator
    Topic
  • #49786
    Robert Milfajt
    Participant

    We have interfaces that go 12 hours or more without sending data.  We are running AIX 5.3 ML 6 and Clovelreaf 5.5rev1.

    We experience connections dropping after 12-14 hours of inactivity.  Since we just upgraded from 3.8.1P/AIX 5.1, we have introduced two new variables.  Our old configuration was stable without these issues.

    The host systems have not changed either, so we cannot point our fingers there, even though the interface log indicates that the host system dropped the connection, even with eo_alias=enable_all.

    Has anyone else experienced this behaviour?  Does anyone have any information on how to use, or experience in using, iptrace or tcpdump to get a trace on the actual socket?

    Any help would be appreciated,

    Bob

    Robert Milfajt
    Northwestern Medicine
    Chicago, IL

Viewing 5 reply threads
  • Author
    Replies
    • #63597
      Tom Rioux
      Participant

      We have experienced something very similar to what you are seeing at your site.  In our case, it was the network that was dropping the connection.  Try talking with your Unix Admin and Network guys.  There may be a setting on there that needs to be tweaked in order for the connection to remain open.

    • #63598
      Michael Hertel
      Participant

      We’ve had issues with 2 hours, not 12.

      But since you’ve upgraded your O/S, the following may be of help.

      http://clovertech.infor.com/viewtopic.php?t=734&highlight=tcpkeepidle

      Have your unix admin check the tcp keep idle parameter in the no settings.

      Hope this helps!

    • #63599
      Chris Williams
      Participant

      We’re on HP-UX 11i,  and the setting is tcp_keepalive_interval in /etc/rc.config.d/nddconf. The default is 7200000 ms (2 hrs).

      The timout can be from any piece of equipment in the route, not just the source or destination systems. We had a router that was dropping the connection.  I changed our Unix keepalive to 90000 ms (15 minutes), and we haven’t had a similar problem since.

      Make the change in the file so it will be effective on the next reboot. Use ndd -c to update the system from the file without having to wait for the next reboot. Check the Unix man pages for more details. Your mileage may vary slightly if you’re running AIX.

      Upgrading your OS probably replaced the file with the network params and caused you to go back to system defaults.

    • #63600
      Robert Milfajt
      Participant

      I figured this one out and of course not a Cloverleaf issue, but an AIX 5.3 issue.  Please refer to the following link for details:  http://www-1.ibm.com/support/docview.wss?uid=isg1IY89429” class=”bbcode_url”>http://www-1.ibm.com/support/docview.wss?uid=isg1IY89429

      I ran the AIX iptrace utility on the particular socket and saw the exact behaviour described in the patch link.  Checking my filesets, I confirmed we had one of the ones affected.

      Basically, the problem is introduced in AIX 5.3 ML 5 and fixed by the time AIX 5.3 ML 6 came out.

      It makes for a great weekend when you can put to bed one of these persnickety varmits.   😆

      Hope this helps,

      Robert Milfajt
      Northwestern Medicine
      Chicago, IL

    • #63601
      Michael Hertel
      Participant

      We are planning on going to ML6.

      Is it fixed in ML6 or do you have to apply the patch?

    • #63602
      Robert Milfajt
      Participant

      ML6 fixes the problem.

      Robert Milfajt
      Northwestern Medicine
      Chicago, IL

Viewing 5 reply threads
  • The forum ‘Cloverleaf’ is closed to new topics and replies.

Forum Statistics

Registered Users
5,043
Forums
28
Topics
9,200
Replies
34,021
Topic Tags
267