[SOLVED] Upgrading from 5.0.16 to 6.0.3 - Tons of bugs

[void]

I am running the network edition (yes, I am not a freeloader and actually paid!) and last week I thought it would be a good idea to upgrade.

I have a server running the free version for personal use, so I tested there first, and the update ran great. No issues at all. I was so happy.

I upgraded my office server last night, from 5.0.16 to 6.0.3_GA_1915 and the upgrade ran without incident.

Now though, I am unable to send or receive any email at all. This server was working just fine, until I ran the upgrade.

Things that the upgrade broke:

1. SSL certificates were gone. I had to go rebuild them, and the Admin UI just kept throwing errors at me so I had to manually provision them from the command line. They work now.

2. Email IS sent, and it is not being received. Any email I send to my server results in a bounce that says:
<nreese@blahblahblah.com>: mail for medata.com loops back to myself


I have compared every field with my working test server to that of my now broke production machine. All I can find is that if I change ANYTHING on the Configuration -> Servers -> mail.blahblahblah.com settings, I get a pop-up with all kinds of interesting Java errors now.

Anybody else seen this? As a commercial user I do have an official ticket open, but it's been two hours without a response so I'm reach out to the community.

Oh also if you are thinking of upgrading to 6 from 5... Might want to not do that unless you have a DR server waiting in the wings to swap in.

[void]

Something must be extremely broken with the update process. I have an engineer at Zimbra, and a Zimbra programmer both logged in to my server right now trying to figure out what is wrong.

Something good has come so far though: by default PostFix is set to soft_bounce=no which is bad. You don't want your postfix problems to cause all your email to bounce immediately do you?

I set this to YES (postconf -e soft_bounce=yes) so that at least for now, while various Zimbra folk troubleshoot the mysterious self-eating server, people sending me email are NOT getting instant bounce messages and will instead queue up on their servers for re-try.

Honestly I would think soft_bounce=yes would be a default, but I can see the logic in letting a user set it for their own needs.

Just wanted to share that little tidbit for the greater good.

[dwill]

The release notes I read stated that versions below 5.0.18 should first upgrade to 5.-.20 then to 6.0.2+. I believe there were some package upgrades which required the intermediary step

[brian]

Originally Posted by dwill

The release notes I read stated that versions below 5.0.18 should first upgrade to 5.-.20 then to 6.0.2+. I believe there were some package upgrades which required the intermediary step

This is an inaccurate statement. The release notes state:

The upgrade path to 6.0.x is ZCS systems at 5.0.2 or later can upgrade directly to 6.0.x.

If you are running a version earlier than 5.0.2, you must first upgrade to 5.0.2 or later before upgrading to ZCS 6.0.

Important: Upgrades from ZCS 5.0.19 or later 5.0.x releases to ZCS 6.0.0 or 6.0.1 will fail. We recommend customers running 5.0.19 or later 5.0.x releases upgrade to 6.02 or later 6.0.x releases when they choose to move to the 6.0.x series.

[brian]

void is an ironic username. The ssl issue was listed as a known issue in the releasenotes for 6.0.3. The postfix problem on your system was a 2 fold problem, you had sendmail enabled and you had filesystem ACL's on /opt/zimbra that was preventing postfix from starting and running properly.

Sorry to hear it took so long for support to pick up your case but glad to hear they resolved the problems you were having.

[dwill]

I concur Brian. Just re-read and cleared that up with myself. Somehere along the lines I was thrashing version numbers against decimals and thought 5.0.2 > 5.0.19. I

[void]

The irony of the username is intentional, I thought was obvious by now.

Support continues to insist the ACLs were causing the issue. Fun part is that the ACls are still there, and were only adding RWX permissions for a junior admin here. They were not taking anything away.

Unless Zimbra has code that detects if a file has an ACL on it, and then holistically shuts down all attempts to access the file, the ACL issue is a red herring.

If it was the ACLs... why is Zimbra working just fine now?

The real issue was the file permissions were all thrashed from the upgrade process and my junior admins bungling with the CHOWN command. I learned something important though: zmfixperms

Also killing that rogue sendmail. Yum must have updated and started it... no idea how that even got there. Junior admin again? Guess who got fired.

Then we rebooted and everything was fine and dandy.

So yah, I still blame bad file permissions/ownership and not ACLs. Believe what you will.