Please share any documentation related to this.
@Prashi_Kapoor it polls every 10 seconds (10000 ms). You can adjust this using the
experimental_pollInterval config option when you setup your
ApolloGateway instance, but please note that you can’t make it less than 10 seconds. If this is a problem, please contact Studio support (email@example.com).
Just as a clarifying note — the
experimental_pollInterval parameter applies to both “managed federation” (which we recommend since it provides a single source of truth for the Gateway via the Apollo Studio registry, backed by its composition and “schema checks” features) as well as “unmanaged federation” (which directly polls and composes from subgraphs).
It’s true that the minimum for managed federation is 10 seconds, but for unmanaged federation, you can set this as low as you’d like, however, it’s still not recommended to have a value that is too low since it can result in schema flapping. Based on the title of your post, the unmanaged federation mode might be what you’re referring to, so I thought it worthwhile to mention this nuance. Overall, we recommend using managed federation so you can “drive” the mechanism of when schema roll-over happens by using the
rover CLI rather than having polling of subgraph service SDL involved at all, which can be volatile during a subgraph rollout if there are multiple instances behind a load-balancer, each on a different version of the schema.
As a similar approach and an alternative to
serviceList, you might consider our new supergraph SDL workflow which is powered by the
rover supergraph compose command – we’ve documented it here.
Thank you for the quick response