Apollo-server: [Federation] Need hot reload of gateway for schema changes in federated service
In Apollo Federation, I am facing this problem:
The gateway needs to be restarted every time we make a change in the schema of any federated service in service list.
I understand that every time a gateway starts and it collects all the schema and aggregates the data graph. But is there a way this can be handled with any plugin or any suggested third party solution?
Can we expect an official solution for this problem very soon?
omkarprabhu-continuum
All 7 comments
Check the Managed Federation:
https://blog.apollographql.com/announcing-managed-federation-265c9f0bc88e
https://www.apollographql.com/docs/platform/federation/
fernandobandeira
on 25 Jul 2019
@fernandobandeira Suppose I do not want any dependency on a third party platform. How can I solve this problem then?
omkarprabhu-continuum
on 26 Jul 2019
I have created a PR that solves this need for non-manged federation: #3117
ronan
on 31 Jul 2019
3102 is also to solve the same problem
hugues33
on 1 Aug 2019
I don't feel polling is the right solution for me and just a question:
Is there a pub/sub way to deal with this?
omkarprabhu-continuum
on 1 Aug 2019
@omkarprabhu-continuum @ronan we don't think that polling downstream services is the right model for a federated gateway. It presents a footgun where gateway flapping (going from one version to another and back while a service is being incrementally rolled back) can happen and/or an intermittent outage of a downstream service can cause composition failures at the gateway. We poll a managed config file for the platform support because that file is updated through a manual hook that teams call via the CLI.
To support teams that can't use the managed version or want to provided additional fallbacks, we are landing full control over the schema fetching processes via observability hooks in this PR https://github.com/apollographql/apollo-server/pull/3110
That PR will allow you to provide your own way to fetch a schema from something like a centralized cache, filesystem, or what have you. If you must call into the downstream resources, that PR will give you the ability to do that as well though we don't recommend it
jbaxleyiii
on 2 Aug 2019
Closing by the many-times-referenced #3110. We'd love to hear any feedback on the custom configuration fetcher and observability hooks!
trevor-scheer
on 4 Sep 2019
Related issues
dbrrt
·
3Comments
stevezau
·
3Comments
Magneticmagnum
·
3Comments
jpcbarros
·
3Comments
jminz
·
3Comments
Most helpful comment
@omkarprabhu-continuum @ronan we don't think that polling downstream services is the right model for a federated gateway. It presents a footgun where gateway flapping (going from one version to another and back while a service is being incrementally rolled back) can happen and/or an intermittent outage of a downstream service can cause composition failures at the gateway. We poll a managed config file for the platform support because that file is updated through a manual hook that teams call via the CLI.
To support teams that can't use the managed version or want to provided additional fallbacks, we are landing full control over the schema fetching processes via observability hooks in this PR https://github.com/apollographql/apollo-server/pull/3110
That PR will allow you to provide your own way to fetch a schema from something like a centralized cache, filesystem, or what have you. If you must call into the downstream resources, that PR will give you the ability to do that as well though we don't recommend it