Amazon ElastiCache for Redis announces improved maintenance update procedures for auto failover enabled clusters. You can now benefit from improved availability during patching, updates and other maintenance related activities that involve planned node replacements. For Redis Cluster configurations that are set up to use Redis Cluster clients, the planned node replacements will now complete while the cluster serves incoming write requests. For non-Redis Cluster configurations, you may notice a brief write interruption, of up to a few seconds, associated with DNS update. 

from Recent Announcements