[SOLVED] Unable to recover from backup

[clemente]

I am currently testing disaster recoveries from previously created backups.

To do this, I installed two identical zimbra instances on different machines. Identical means: Both machines have the same ubuntu versions and same zimbra versions. The "productive" instance lives alone on a physical machine, the "recovery" instance lives within a vm.

I backed up the productive zimbra according the suggestions in the wiki page "Open Source Edition Backup Procedure". I can easy and successfull recover a backup up on the same (productive or recovery) machine. This is: I can recover a backup taken yesterday and switch to yesterdays state.

According to the mentioned wiki site, the backup contains the complete /opt/zimbra-content and recover means: I replace the current /opt/zimbra with the backed up content.

If recovering a productive backup on the recovery machine, I run into trouble. The recovery machine starts and claims to run normal. Processes are up and running but no web frontend is responding.
If accessing user or admin page the server does not send any data to my browser until the browser runs into a timeout.

Both machines have same hostnames and same /etc/hosts files. To be on the secure side, I configured a dns for the recovery machine as well as for the machine I use to open the zimbra pages to make sure that the hostname resolves to the correct (the recovery machines) ip address.

I checkes the backups using md5-sum to be sure that nothing bad happened while transfering the backup.

I see no error, empty log files and no page content.

Does anyone point me a direction where to search (or better find...) the problem?

Thanks a lot,
Clemens

[phoenix]

I assume that your 'backup' machine has the same FQDN as the original server and there are DNS A & MX records pointing to it? What do you see in the log files, are you sure all the services are up (post the output of a 'zmcontrol status')?

[clemente]

Thank you for your reply!

Yes, both machines have the same fqdn.
The DNS configured for my test scenario has mx and a records configured and I checked the configuration with dig.

zmcontrol status returns following (I just replaced the domain name) :

zimbra@zimbra:~$ zmcontrol status
Host zimbra.ANONYMIZED.de
	antispam                Running
	antivirus               Running
	ldap                    Running
	logger                  Running
	mailbox                 Running
	mta                     Running
	snmp                    Running
	spell                   Running
	stats                   Running

I am not sure what log file is really interesting...
- /opt/zimbra/jetty/logs: No current logfile - just old log files reflecting actions before recovering the backup.
- /opt/zimbra/log: Same as Jetty's logs.
- /var/log/zimbra.log: Contains some server-startup-lines (amavix, clamd, postfix, etc.) but nothing that looks bad for me.

Does this ring any bell I cannot hear? Or do you need more information?

Thanks a lot, Clemens

[phoenix]

Did you run the zmfixperms script on the 'backup' server after moving the /opt/zimbra directory?

[clemente]

Yes, I did.

[clemente]

I think I did find a problem...

There's something weird with Jetty's configuration.
I will post as soon as I know more...

[clemente]

Well - I found the problem - ZImbra is cool and I am stupid...

The short story:
I did not make sure that the system user "zimbra" has the same numeric id on the recovery machine.

The long story (because I worked long time on this issue...):
I somehow crashed the clock on the 'recovery' machine nad lived one month in the past. Considering the logfile's timestamps I did not recognize that there were fresh logfiles.
In one file I found some IOExceptions thrown by jetty and complaining that jetty cannot open several files.
This led me to the fact that Jetty runs as an user unknown to the system - what looks really weird to me. According to $(ps -faux) Jetty runs with user 1004. This id has no entry in /etc/passwd. I did not expected Linux to allow this - I will do some more investigation on this. ;-)
Comparing productive and recovery machines I found that the zimbra user has different ids.
After changing the id on my recovery machine to the id on the productive machine (in fact user and group id), the server starts up and responds fine.

Thanks a lot,
Clemens


Thanks to everyone reading and thinking about this issue!