Google Cloud's managed instance group rolling updates let you control update behavior with two parameters: maxUnavailable (how many instances can be down/unavailable at once) and maxSurge (how many extra instances above the target size can be created temporarily). To guarantee capacity never drops during the rollout, you must set maxUnavailable to 0 (no instances go offline before their replacement is ready) and maxSurge to a value greater than 0, such as 1, so new instances are created and become healthy before old ones are removed. This combination ensures capacity stays at or above the original level throughout the update, satisfying the requirement that available capacity not decrease.
Community Discussion