Forum Discussion

Hamish's avatar
Hamish
Icon for Cirrocumulus rankCirrocumulus
Mar 08, 2012

v10.2.3 units go active on boot

Just wondering if anybody else noticed when rebooting their BigIP 10.2.3 units in an HA pair, one of the boxes always goes active after reboot before the system has a chance to replicate the connection table from the partner unit?

 

 

I thought it was isolated and just something bad on a pair of 1600's... But after upgrading a pair of 6400's (10.2.3HF1), I notice the same thing.

 

 

Sadly enabling debugging on sod 'fixes' it... At the expense you have to put up with a truckload of verbose logging.

 

 

Anyone else?

 

 

H

 

config
design
  • lkchen's avatar
    lkchen
    Icon for Nimbostratus rankNimbostratus
    Apr 28, 2012
    Yeah, I saw that when I upgraded our 6400's to 10.2.3 and later 10.2.3HF1.

     

     

    Was annoying, because I was upgrading the standby, and didn't notice that it was active while I was still working on getting the upgrade configs working on it. (had upgraded from 9.3.1, so got bit by change in behavior of HTTP monitors and the smbclient monitor being broken.) So, services were down...even though they weren't.

     

     

    Made the mistake of thinking I could upgrade the standby ahead of the scheduled maintenance window with no disruption.

     

     

    Wonder if that's the sod fix that is in 10.2.4?

     

     

    When I upgraded from 9.3.1 to 10.2.3....the 10.2.3 took over when it shouldn't have. But, when I was upgrading to 10.2.3HF1...I ended up with an active/active situation, and had to force one out.

     

     

    L
  • Beinhard_8950's avatar
    Beinhard_8950
    Icon for Nimbostratus rankNimbostratus
    May 03, 2012
    I have not seen this on the 11050 boxes.

    How is your setup?

     

     

     

    For network failover I use mgmt and unicast.

     

     

     

    However, I will verify this in the lab in some days because I know that I had one unexpected failover in the past but that was when I struggled with some cmp problems as well.

     

     

     

    /Beinhard

     

  • lkchen's avatar
    lkchen
    Icon for Nimbostratus rankNimbostratus
    May 08, 2012
    Only serial failover. Though we have vlan failsafe configured for one external vlan and one internal vlan.
  • Beinhard_8950's avatar
    Beinhard_8950
    Icon for Nimbostratus rankNimbostratus
    Jun 04, 2012
    i have checked now in my production boxen, however I´m not using serial but to avoid this behavior (without serial ) it´s mandatory to have unicast for mgmt interface (and for redundancy on some more vlan) and then it´s OK.

     

     

     

    /beinhard

     

     

  • Hamish's avatar
    Hamish
    Icon for Cirrocumulus rankCirrocumulus
    Jun 05, 2012
    FWIW I only see this on 6400's and 1600's. Not 3400's. That's with network failover only. No serial. It's consistent, and the 1600 that does it has just been reinstalled with a disk format. Same issue.

     

     

    Sadly they're in production so the opportunities to run diagnostics is pretty small...

     

     

    H
  • giltjr's avatar
    giltjr
    Icon for Nimbostratus rankNimbostratus
    Jul 25, 2012
    A little late, but I was under the impression that you could not HA between different versions of LTM. I thought I read that in the doc dealing with doing upgrades on a HA pair. That if you were going to upgrade the standby box, you had to disable/disconnect all network connections.

     

     

    I know when I did our upgrade from 9.4.4 to 10.2.3 (6400's) that it was I did and what was recommended to us by F5. We are doing network fail over.