We have already discussed maintenance in the previous sections. What we never covered was self-imposed archival delay. If we're performing maintenance or the primary node crashes, it is a very good idea to either delete the /etc/cron.d/sync_archive script or comment out the rsync command itself until the maintenance is complete. This hour-long barrier helps avoid propagating corrupt data, but there's no reason to take excess risks.
Some environments have another pair of servers in a different data center that acts as disaster recovery. If this is our setup, any running server on the disaster-recovery side should be stopped while we modify or rebuild the primary or secondary servers. The reasoning is the same: if there is a ...
