- 최신
- 최다 투표
- 가장 많은 댓글
Targets are defined dynamically by the unique ClientId parameter used to call the GetConfiguration API (https://docs.aws.amazon.com/appconfig/2019-10-09/APIReference/API_GetConfiguration.html).
During a deployment, we probabilistically assign the ClientId to a percentage bucket, 0-99. When the percentage complete for a deployment passes the percentage bucket the ClientId was assigned, it receives the new deployment.
For example, say you use a linear deployment strategy with a deployment duration of 5 minutes and a growth factor of 10. If you have 100 targets, then roughly 10 targets will receive the new deployment every 30 second interval.
Thanks. That's very helpful. Is it possible that the same clientId might be assigned to a different percentage bucket in different deployments even for the same deployment strategy?
We don't make any guarantees about when a specific target will receive a deployment. The behavior will be stable within a deployment, but you shouldn't make any assumptions that ordering in one deployment will be similar in the next.
Is this still how the targets are selected, given the API has changed to recommend StartConfigurationSession now?
Secondary question: how are the targets defined, are they the end user/caller of say an API/Microservice or the instance of the AppConfigClient used to retrieve configuration by the API/Microservice