Problem Upgrading from 4.0.1 to 4.0.x

[shankwc]

banshee.XXXXX.com has address 10.2.3.137

I feel like i'm going in circles. I'm documenting this here for my own sake as well. I was running 4.0.1 but when I upgrade to 4.0.2 I get the INVALID_PDU problem trying to connect. Trying to go from 4.0.2 -> 4.0.3 fails with :

This appears to be 4.0.2_GA
Schema upgrade required
Sun Nov 19 10:32:43 2006: Running /opt/zimbra/libexec/scripts/migrate-ComboUpdater.pl
Sun Nov 19 10:32:45 2006: Verified schema version 26.
ERROR 1060 (42S21) at line 1: Duplicate column name 'contact_count'
Error while running '/opt/zimbra/bin/mysql --user=zimbra --password=jbPc95sRbjXlbk9BJRt1sGva --database=zimbra --batch --skip-column-names'. at /opt/zimbra/libexec/scripts/Migrate.pm line 136, 
 line 287.
Sun Nov 19 10:32:45 2006: Script failed with code 1 - exiting
UPGRADE FAILED - exiting

I applied the script you suggested (which worked on 4.0.1 taking it 4.0.2 BTW) but fails like this now:

banshee:~/zcs# /opt/zimbra/libexec/zmsetup.pl
Operations logged to /tmp/zmsetup.log.1270
Setting defaults...Setting defaults from existing config...Done
Upgrading from 4.0.2_GA_362 to 4.0.3_GA_406
Sun Nov 19 10:37:30 2006: Stopping zimbra services
Sun Nov 19 10:37:41 2006: Starting mysql
Sun Nov 19 10:37:43 2006: Checking logger mysql status
Sun Nov 19 10:37:44 2006: Starting logger mysql
Sun Nov 19 10:37:44 2006: SELECT value FROM config WHERE name = 'db.version'
This appears to be 4.0.2_GA
Schema upgrade required
Sun Nov 19 10:37:44 2006: Running /opt/zimbra/libexec/scripts/migrate-ComboUpdater.pl
Sun Nov 19 10:37:46 2006: Verified schema version 26.
ERROR 1060 (42S21) at line 1: Duplicate column name 'contact_count'
Error while running '/opt/zimbra/bin/mysql --user=zimbra --password=jbPc95sRbjXlbk9BJRt1sGva --database=zimbra --batch --skip-column-names'. at /opt/zimbra/libexec/scripts/Migrate.pm line 136, 
 line 287.
Sun Nov 19 10:37:47 2006: Script failed with code 1 - exiting
UPGRADE FAILED - exiting

Could the mysql password be out of sync? I'm stuck and frustrated. I was up late working on this and barely slept. I had a backup prior to doing all this and I rolled back my install when I saw the PDU and 4.0.3 upgrade error. It was running ok when I left but at 2am we shutdown zimbra for an rsync. After that it doesn't start back up - stops cold at LDAP starting. I tried using db_recover but it says the files are for a different version and fails to correct them. So now I'm in real trouble because my backup copy doesn't work and trying to pull this from off-site backup will take hours upon hours for the 40G of email.

So I really appreciate your help. Thank you.

[shankwc]

also - no firewall problems. we have a border firewall that port forwards. when i first started with zimbra i tried to firewall it all but there was so many services it was a PITA to get it to work. So I just left it up to the border FW.

[shankwc]

Ok, I've gotten a little farther. At this point I'm running v4.0.4. I was able to get the upgrade to run by manually renaming the column contact_count in mysql table mailbox. This must have been modified when I first attempted to upgrade to 4.0.3 and it failed (on mysql errors).

Now I'm trying to solve the login problem and I suspect it is a port issue caused most likely by an name resolution problem.

I am NOT able to telnet to ldap (389) on the localhost loopback ip but I am able to do this on the regular exeternal ip. Which of these should LDAP be listening on? How can I get it set on both?

Thanks

[shankwc]

Well I'm not any further along. I've confirmed that there is nothing blocking any ports by moving aside my primary zimbra dir and reinstalling a vanilla copy. I am certain this is a name resolving issue between my internal and external names but i just can't narrow it down.

Is there any way to move the accounts from the broken install into the vanilla install without also moving the bad configs or libraries?

[shankwc]

UPDATE: I gave up and rolled back to the 4.0.1 version. Scary that I could never find out what happened. The only other reports of this issue seem to stem from schema changes. I think that there was a significant schema change going to 4.0.2 and I don't think my system took to the change well.

Are there any tools to check the ldap and mysql schemas and verify they are correct?