I’ve spent two nights upgrading Zimbra 4.5.11 on 32 bit Centos 4 to Zimbra 6.0.3 on 64 bit CentOS 5. The upgrade process didn’t cost too much time, but the backup process was killing me. First rule of system administration “Never do anything to production system without taking a backup first”. I was forced to do it at night because people are using it at working hour. Second rule of system administration “keep the downtime as minimum as possible especially when many people are using the system”.
First night, I upgraded Zimbra from version 4.5.11 to 6.0.3 in place. To do so I can’t upgrade it directly, the upgrade path forced me to upgrade it twice from 4.5.11 to 5.0.20 first, and then from 5.0.20 to 6.0.3 or else the upgrade process will fail and all your data will be fucked up. I didn’t move it to the new server yet because I felt very sleepy already. In the morning, most users wouldn’t notice until they realized, there are many new feature on their webmail, it’s appearance looks different, it looks good!.
Second night, I migrated Zimbra 6.0.3 from 32 bit Centos 4 to another machine, 64 bit CentOS 5. It only needed one upgrade step, but once again, remember the first rule of system administration. At least it was quicker than the first night, nothing wrong happened and Zimbra successfully migrated. I sent notification to all users to accept the new SSL certificate if they are using secure connection. In the morning, once again people wouldn’t notice the different.
Now Zimbra has been upgraded to the newest version and migrated to new machine, everything was done at night, when they were sleeping. The only thing they will notice is, two days in a rows they saw me came to the office after lunch time, and maybe they are thinking that I will be sacked soon.