Terraform module for a Redis MemoryStore in GCP
This upgrade recommends using Google VPC's PRIVATE_SERVICE_ACCESS connection mode. This upgrade will cause the redis instance to be allocated a new IP CIDR range (as per the connection mode) and, hence, requires the redis instance to be destroyed and then re-created.
If you have setup a DNS address for the redis instance using var.dns_zone_name (available since v2.1.0), and if your applications are already using the redis instance's DNS address instead of its IP address to connect to it, then it should ideally be okay to proceed with this recommended upgrade. Notice, however, that as this will cause the redis instance to be destroyed and then re-created, the redis instance will certainly be unavailable during the upgrade (usually 10 minutes). You can minizmie potential customer impact during this upgrade process by letting your application temporarily bypass the redis cache and call your databases directly for this brief period of time.
If, however, you would like to avoid all these preparationss & arrangements, simply set the variable use_private_g_services = false when upgrading - which uses Google VPC's DIRECT_PEERING connection mode - which is what has been used by default prior to this version.
Reference: To learn more about these 2 connection modes used by a Redis memorystore instance, visit here.