Primary portal machine post-upgrade successful, standby fails (10.6.1)

756
6
11-29-2018 12:09 PM
MichaelSchoelen
Occasional Contributor III

When upgrading a highly available portal in our disaster recovery environment, we installed 10.6.1 on both portals successfully. Then on the primary machine, we hit the Complete Upgrade button, which was also successful. 

Now we are seeing that the standby machine did not complete the post-upgrade steps. 

Hitting upgrade on the standby machine shows this message (domain.com covers our actual domain):

"Upgrade failed. You can only upgrade the primary portal machine domain.com when portal is in a highly available configuration."

Interesting that it say our load balancer DNS "domain.com" rather than the primary machine name. 

Is there a way to force the standby upgrade to complete?

0 Kudos
6 Replies
JonathanQuinn
Esri Notable Contributor

Do you remember if you installed the software on the primary machine before the standby? If you installed Portal on the primary first, then that would cause a failover and promote the standby to primary. When you want to continue the upgrade on the original primary, you'll receive the error you mentioned. The error you see is strange, though, we should be grabbing the primary machine name and telling you that you must upgrade that machine.

Have you tried to upgrade the other machine instead?

0 Kudos
MichaelSchoelen
Occasional Contributor III

We followed these steps:

  1. Uninstall 10.5.1 Portal Web Adaptors (success)
  2. Run 10.6.1 Portal installer on standby machine (success)
  3. Authorize standby (success)
  4. Run Portal installer on primary (success)
  5. Authorize (success)
  6. Run Continue upgrade on Primary (It said it was successful)
  7. Attempted upgrade on the standby (error)

So this is in our standby environment, so we have a modified etc/hosts file that is named "domain.com"-- I wonder if that was the problem.

0 Kudos
JonathanQuinn
Esri Notable Contributor

When did you add the "domain.com" etc\hosts entry? What was the reason for doing so? The Portal will pick up that hostname as it's own hostname since the IP address resolves to the same machine, which is why you're seeing the machine name listed in the error.

0 Kudos
MichaelSchoelen
Occasional Contributor III

It was added pre-upgrade. We have that in place to allow WEBGISDR to run in a disconnected environment.

That makes sense--is there a way to comment out that line in the hosts file then re-run the upgrade procedure?

0 Kudos
JonathanQuinn
Esri Notable Contributor

Is this a VM? I'd create a snapshot first. We backup directories during the upgrade but it's a good practice just in case.

But yes, try to stop Portal, uncomment that line, start Portal, and upgrade. I've never tried that so I don't know what will happen, but it's worth a shot. If that fails, uninstall and reinstall with the line uncommented and try again.

Can you describe why you needed that value for the DR tool to run?

0 Kudos
wadsonmakari
New Contributor III

Did you ever find a solution to this issue? Is there property on the primary portal machine that one can check to see if it has HA configuration set correctly?

0 Kudos