Generally worked as expected, but there is one case I found out causing potential issues: if new revision need to spin up, but unable to succeed, the killed healthy old revision won’t be coming back. With availability strategy, we won’t be in this situation, the worst case is old and new revisions with traffic split. The resourceUtil strategy is something we should use and mark with use with caution
The principle for the resourceUtil strategy is scaling down the old revisions first and then scaling up the new revisions. We cannot violate the rule by far, so this issue will remain being discussed.
Generally worked as expected, but there is one case I found out causing potential issues: if new revision need to spin up, but unable to succeed, the killed healthy old revision won’t be coming back. With availability strategy, we won’t be in this situation, the worst case is old and new revisions with traffic split. The resourceUtil strategy is something we should use and mark with use with caution