5.0.2 RHEL4 -> 5.0.16 RHEL5 upgrade, lost local config

[dmenden]

I've been running the Zimbra open source version 5.0.2 on CentOS 4 for about a year, and yesterday tried to update the OS and Zimbra. I upgraded to CentOS 5.3, then tried to run the 5.0.16 RHEL 5 installer. Everything looked fine but it stopped responding for over an hour when trying to update the logger schema. I killed the process, did an rpm -e on the logger package, and tried again. It hung again trying to start mysql. Killed it and tried to install the old version. It fails when trying to backup the ldap server. Tried to install the new version. The packages are now all installed at level 5.0.16, but the installer still fails with:

Backing up the ldap database...failed.

slapcat: bad configuration file!

Trying to start zimbra gives this error:

$ zmcontrol start
Host localhost
Unable to determine enabled services from ldap.
Enabled services read from cache. Service list may be inaccurate.
        Starting ldap...Done.
FAILED
ldap_url and ldap_master_url cannot be the same on an ldap replica

zmlocalconfig appears to have default values. It looks like all the data is still in the db, openldap, and store directories, but Zimbra doesn't know how to access them. There are old config files in the .saveconfig directory, but an install.sh -r /opt/zimbra/.saveconfig/5.0.2_GA_1975.RHEL4-20080130212006/2/config.save just gives the same "slapcat: bad configuration file!" error. Any ideas on what I can do to recover the configuration?

[uxbod]

Welcome

By any chance are you on a private IP ?

cat /etc/hosts
cat /etc/resolv.conf
dig yourdomain mx
dig yourdomain any
host `hostname` <- note back ticks and not single quotes (copy and paste)

[dmenden]

Nope, public IP. Host files and DNS are all okay, other than changing the MX record today to point to another host while this one is down.

[phoenix]

Nope, public IP.

Is it behind a firewall? How about posting the diagnostic information you were asked for?

[dmenden]

No firewall. What are you thinking might be the issue here? Is there some port you think might be blocked?

[root@mail ~]# cat /etc/hosts
127.0.0.1               localhost.localdomain localhost
my.ip.add.149           mail.mydomain.org

[root@mail ~]# dig mail.mydomain.org any
;; ANSWER SECTION:
mail.mydomain.org.        3600    IN      A       my.ip.add.149
mail.mydomain.org.        3600    IN      MX      10 mail.mydomain.org.

[root@mail ~]# host mail.mydomain.org
mail.mydomain.org has address my.ip.add.149
mail.mydomain.org mail is handled by 10 mail.mydomain.org.

[phoenix]

No firewall. What are you thinking might be the issue here? Is there some port you think might be blocked?

[CODE][root@mail ~]# cat /etc/hosts
127.0.0.1 localhost.localdomain localhost
my.ip.add.149 mail.mydomain.org

This is incorrect, the format should be as follows:

127.0.0.1               localhost.localdomain localhost
my.ip.add.149           mail.mydomain.org mail

[root@mail ~]# dig mail.mydomain.org any
;; ANSWER SECTION:
mail.mydomain.org.        3600    IN      A       my.ip.add.149
mail.mydomain.org.        3600    IN      MX      10 mail.mydomain.org.

This is not the command I asked you to run, I want the output of the following:

dig yourdomain.com any
dig yourdomain.com mx

using your domain name and not the FQDN of your server.

[root@mail ~]# host mail.mydomain.org
mail.mydomain.org has address my.ip.add.149
mail.mydomain.org mail is handled by 10 mail.mydomain.org.

That's also not the command I asked you for, I want you to run the exact command I gave you with no changes (i.e. do not add your hostname).

[dmenden]

That's also not the command I asked you for, I want you to run the exact command I gave you with no changes (i.e. do not add your hostname).

[root@mail ~]# host `hostname`
mail.mydomain.org has address my.ip.add.149
mail.mydomain.org mail is handled by 10 mail.mydomain.org.

The domain name is mail.mydomain.org. That is also the hostname. That is also where the MX record points. Mail goes to user@mail.mydomain.org. None of that changed with the upgrade.

May I ask how one would go about recovering the local config? What would I do differently if I did find something wrong with the name lookups?

[phoenix]

The domain name is mail.mydomain.org.

That's a subdomain.

That is also the hostname. That is also where the MX record points.

That's incorrect, the domain name should be mydomain.org and the A & MX records point to the FQDN of the server i.e. mail.mydomain.org and the public IP address.

[dmenden]

That's a subdomain.
That's incorrect, the domain name should be mydomain.org and the A & MX records point to the FQDN of the server i.e. mail.mydomain.org and the public IP address.

I'm sorry, but it is correct. Yes, mail.mydomain.org is a subdomain of mydomain.org, but it is also the one and only domain name of this machine. mydomain.org has nothing to do with this installation. Neither the A nor MX records of mydomain.org point to this IP. Mail for mydomain.org is handled by a different machine.

Can you tell me what difference it might make in restoring my local config? As I've said, nothing about the naming was changed during the upgrade. I've run upgrades previously with this naming. Assuming all of those commands came back with the values you think they should, what would I do next?

If I were to move all the data to another directory, do a fresh install of 5.0.2, and then move the data back into place, is there any way to reload the local config from backup? I think the problem might just be that Zimbra doesn't know how to log into the local ldap and mysql server. When I ran the installer, it never made it to the configuration menu, and now zmlocalconfig doesn't have any host or password information. That information is available in the saved config files.