Forum Discussion

dundovic_25174's avatar
dundovic_25174
Icon for Nimbostratus rankNimbostratus
Apr 22, 2013

BigIP LTM 10.2.3 -> 11.3.0 upgrade - configuration issues

I'd like to know if anybody here had issues upgrading LTM from 10.x.x to 11.x.x. This was my 3rd attempt, in all three cases ended up calling F5 support, but unfortunately, even with their assistance not being able to have a success. My first attempt was with 11.0 release, then 11.2 and last week, I gave it another chance and tried to upgrade to 11.3. I was following documentation provided on F5 support site:

 

http://support.f5.com/kb/en-us/products/big-ip_ltm/manuals/product/bigip-upgrade-active-standby-11-2-0/1.html

 

I started with standby box, following the documented process, no errors/warnings reported, but when it got rebooted to 11.3, none of the config entries were present (this possibility is not even outlined in the documentation). Basically, node seemed like it's fresh installed, not upgraded. To resolve this, I imported previously backed up 10.x config. This helped bringing the config back, but not to the full extent. The node entry fields were messed up. Within the 'name' field there were IP addresses and node names were pushed to the 'description' field. i contacted F5 support and was told that this is due to some crucial changes within 11.x system release and that all of the nodes have to be manually reconfigured again(!?). This wouldn't be a big issue with a few nodes, but there's a couple of hunred nodes already defined. Manual re-entry is definitely not an option. I didn't even go further to check if there's any discrepancy on pool/vip level or any other config component. I had to roll-back again as this didn't make any sense.

 

It's really hard for me to apprehend that this kind of upgrade issue should be considered acceptable. F5 documentation and KB doesn't have a word on this kind of situation and doesn't state anything about 10.x/11.x config compatibility. Is there anybody out there who was performing similar upgrade without any issues or anybody who bumped into this (or any) kind of problem and was able to fix it?

 

Regards,

 

Sasha

 

management
monitoring

5 Replies

Sort By
  • Tim_K_92675's avatar
    Tim_K_92675
    Icon for Cirrostratus rankCirrostratus
    Apr 22, 2013

    We've been struggling with the 10.2 to 11.2 upgrade since last September.

     

     

    1. We had a .NET issue manifest itself which resulted in us having to implement an enteprise-wide system.net configuration update.

     

    2. We also noticed the node name/description switch as well. I'm testing an automated process to re-customize the name field values.

     

    3. We use multiple partitions/route domains per LTM and use a custom icmp monitor configuration as opposed to using a default one. The issue is that the custom monitor as defined in 10.2 does not include the partition in the path name (of course), and it was not automatically updated during the upgrade. This resulted in all nodes being marked as down because the path to the icmp monitor could not be found. We have been unable to modify the node config afterwards...not in the GUI (the node properties would not even display), nor via CLI, nor by manually editing the bigip.conf file directly. Everything we tried complained about not finding the path to the monitor during some sort of validation process. This complete failure occurred in our passive production data center, so we were able to roll back without affecting anything major. However, our QA environment has started to have this happen when someone disables a pool member or tries to make any sort of change to a node property...the node becomes completely unusable nor updateable. Why it didn't have a complete failure of all nodes like it did in our passive production environment has yet to be understood. We've been working with support since last week about this.

     

     

    I spoke with F5 support this morning and their product development has reproduced this monitor issue in 11.2 and 11.3. I hope to have their workaround in hand later today. I would assume a patch would also be forthcoming.

     

  • dundovic_25174's avatar
    dundovic_25174
    Icon for Nimbostratus rankNimbostratus
    Apr 22, 2013
    Thanks for giving some more input on this topic, Tim! Maybe this thread should be a way to collect and expose all of the recognized upgrade issues and share between the F5 community. I hope somebody from official F5 support has an eye on it.
  • afedden_1985's avatar
    afedden_1985
    Icon for Cirrus rankCirrus
    Apr 22, 2013
    I have run into this same issue > rebooted to 11.3, none of the config entries were present

     

    to resolve this support told me to run this command to reload the v10 config into v11. Run this command when you don't see what you should see in the gui>> /usr/libexec/bigpipe daol

     

    FYI>Note in this directory /config/bigpipe , all the original v10 configs exist and the daol cmd uses them to reinstall the config in v11 format

     

    This command restored all my virtuals and pool that were misisng before running that command.

     

    after that I had to fix an audit error in the bigip_base.conf since audit logging was set to verbose in v10 and verbose won't convert to v11 unless in V10 is set to enabled or disabled or fixed after the v11 upgrade with vi

     

    we had some WAM issues where we had to correct the http compression profile and the http acceleration profiles.

     

    the next place we had issues was building the Trusts but we have a procedure that works to correct that now.

     

    good luck
  • dundovic_25174's avatar
    dundovic_25174
    Icon for Nimbostratus rankNimbostratus
    Apr 22, 2013
    I had to run the same command in order to restore config, but then, I started noticing those discrepancies in node definitions. I didn't even go that far to check the rest of the config items before rolling back. What really bothers me is the fact that F5 doesn't say a word about 10.x/11.x incompatibility and doesn't provide any additional hint/articles on possible issues that may occur during the upgrade process. Another issue is that once the upgrade process is completed, there's no sign of report/warning that would tell end user about config not being processed at all. Looking at their 'upgrade howto' document, one would say upgrade is an absolutely straightforward process. Worst of all, even after three subreleases, they still didn't address this problem.....
  • Hamish's avatar
    Hamish
    Icon for Cirrocumulus rankCirrocumulus
    May 14, 2013
    I migrated all mine by hand. It was a good chance to ensure that the standards for naming schemes etc were re-adhered to... Plus you get to test along-side which keeps the support, management and users happier (Because you find the issues before they hit live production). [Mind you I was also moving from 2 node ACA to 6 node clusters and APM and a few other bits & pieces]

     

     

    H