Reducing downtime during migration to new server
I am planning to migrate from 4.5.11 SLES9 32-bit to 4.5.11 CentOS 64-bit. I am following the directions here:
Network Edition: Moving from 32-bit to 64-bit Server - Zimbra :: Wiki
These instructions have the old server offline for quite a while during the copy of /opt/zimbra/store/* and index/*. In my environment, this copy can take several hours.
One thought was that I could try this:
1. set up ZCS on the new server per the above directions and then stop ZCS on the new server
2. do an rsync from the old server's store and index to the new server's store and index while the old is still online and servicing users
3. stop ZCS on the old server
4. do another rsync (this one going much faster since it would only have to move the changes that occurred since the last rsync)
5. continue with the migration (flip the IP addresses, import the LDAP, etc....) and bring the new server online
This way, I could move 99.9% of the store data while the old server is still online.
What are the potential pitfalls of this approach?