Booting wrong version on cf giveback

[kelwood1976]

I upgraded my 3240 HA system from 8.0.2 to 8.1 with some trouble and now one node will not automatically boot the correct version when testing the HA abilities.  Example: fillerA takes over and everything works.  filerA does a giveback to filerB and filerB trys to boot 8.0.2 and PANICs.  After the failure, at the LOADER prompt on filerB I can do a boot_backup and it will boot 8.1. 

A take over and giveback from filerB works fine.  filerA will boot with 8.1 on giveback. 

How do I get filerB to boot 8.1?  I'd like to get this fixed without an outage if possible.

KW

[jimmcclung]

I have the same issue except that I have a V3270 cluster and I am trying to upgrade from 8.0.2P2 to 8.0.3.

NetApp support worked with me for several hours and found that a environmental variable will have to be set in the USB flash via booting to a netboot image. I will have to open another case and support session to accomplish this task as I ran out of time on my previous maintenance window.

Update this if you find an easier workaround.

[scottgelb]

Definitely open a support case on this.  The boot image didn't get updated as primary.  I have also seen this happen on an NDU upgrade where someone does a "cf takeover -f" instead of cf takeover... forced takeover doesn't swap the image (which is a reason the upgrade advisor has a step to terminate cifs before a normal takeover).  It sounds like a normal takeover did not occur... so the node needs ONTAP pushed to it and either a normal takeover or reboot of that node.  But I would make sure support is guiding you on that for a production system.  They may also be able to talk you through setting the 8.1 image as primary instead of pushing ontap and will have the best resolution that is supported.

[mbonneau79]

I had this happen recently as well during a 7.3.2 to 8.0.2 upgrade. Failed first controller over to second, performed the update process. First controller came up with new version, all good. CIFS wouldn't terminate on the second controller so I used "cf takeover -f" and performed the upgrade. It appeared to upgrade fine but after reboot it was running the wrong version.

I ended up netbooting the second controller and pointing it to the DOT8 upgrade image - that worked fine (in the end it did, anyways. IBM swore I had a bad CF card and made me wait until the following day to keep troubleshooting).

[scottgelb]

The cf takeover –f was the same issue there… When we ran into it at a customer, support came back and confirmed that the takeover –f does not swap the image…so the new image stays as the backup image (not marked as primary) unless a normal takeover or reboot occurs after the ontap download.

[aborzenkov]

Could you elaborate on how you fixed it (more detailed steps)? In particular, what "pointing it to the DOT8 upgrade image" means?

[mbonneau79]

Sure, it's pretty simple. Netboot the filer using the appropriate netboot image. In my situation, I had originally tried to upgrade using the 7.3.x to 8.0.2 image, so when I tried to use it from the netboot image it failed. I had to go grab the DOT8 upgrade image for 8.0.2, stick it on a web server, and run through the netboot process again, then it worked.

[aborzenkov]

Was it enough to just boot until special menu? Because normally you cannot fully netboot into running Data ONTAP with 8.x; you stop in special menu. Did you do anything in special menu?

[christin]

I'm a bit late to chime in, but this indeed sounds like a support related question. If you haven't already done so, please open a case NetApp Technical Support.

Regards,

Christine

[kelwood1976 · Accepted solution]

Support had me do another "software update 81_q_image.tgz -r" on filerB.  I figured this was the fix but was happy to hear support agreed!  What it did though was download 8.1 to the 2nd image thus putting 8.1 software image on both primary and backup boot images.  This wasn't a big deal to me.  Tested failover and giveback and all worked.