July 16, 2026

@graphql-mesh/fusion-runtime@1.11.0

Minor Changes

• #2449 65ecf28 Thanks @advoretsky! - Add opt-in graceful supergraph reload ("generation overlap").

  By default, when the supergraph reloads, the previous generation (executor +
  subgraph transports) is disposed immediately, aborting any in-flight operation
  on it with a SCHEMA_RELOAD / 503 error (queries are retried on the new schema;
  mutations are not).

  With the new gracefulSchemaReload config the previous generation is kept alive
  and only new requests are routed to the new one, so in-flight queries and
  mutations finish on the schema they were admitted under. Operations are
  reference-counted per generation for their whole lifetime (across all subgraph
  hops, and until @defer/@stream streams end). A superseded generation is
  disposed once idle, or force-disposed after drainTimeout; maxConcurrentGenerations
  caps how many generations may overlap. Subscriptions are not pinned: one on a
  superseded generation ends with SCHEMA_RELOAD when that generation is
  disposed — immediately on reload when nothing is draining, otherwise once the
  last in-flight operation finishes (at the latest after drainTimeout) — and
  the client then reconnects against the new schema. Disabled by default.

  export const gatewayConfig = defineConfig({
    gracefulSchemaReload: { drainTimeout: 10_000 }, // ms; default off
  });

@graphql-hive/gateway@2.10.5

Patch Changes

• Updated dependencies [65ecf28]:
  • @graphql-hive/gateway-runtime@2.10.0
  • @graphql-hive/plugin-aws-sigv4@2.0.54
  • @graphql-hive/plugin-opentelemetry@1.4.36
  • @graphql-mesh/plugin-prometheus@2.1.52

@graphql-hive/nestjs@2.0.85

Patch Changes

• Updated dependencies []:
  • @graphql-hive/gateway@2.10.5

@graphql-hive/plugin-aws-sigv4@2.0.54

Patch Changes

• Updated dependencies [65ecf28]:
  • @graphql-mesh/fusion-runtime@1.11.0

@graphql-hive/plugin-opentelemetry@1.4.36

Patch Changes

• Updated dependencies [65ecf28]:
  • @graphql-hive/gateway-runtime@2.10.0

@graphql-mesh/plugin-prometheus@2.1.52

Patch Changes

• Updated dependencies [65ecf28]:
  • @graphql-hive/gateway-runtime@2.10.0

@graphql-hive/router-runtime@1.4.16

Patch Changes

• Updated dependencies [65ecf28]:
  • @graphql-mesh/fusion-runtime@1.11.0

@graphql-hive/gateway-runtime@2.10.0

Minor Changes

• #2449 65ecf28 Thanks @advoretsky! - Add opt-in graceful supergraph reload ("generation overlap").

  By default, when the supergraph reloads, the previous generation (executor +
  subgraph transports) is disposed immediately, aborting any in-flight operation
  on it with a SCHEMA_RELOAD / 503 error (queries are retried on the new schema;
  mutations are not).

  With the new gracefulSchemaReload config the previous generation is kept alive
  and only new requests are routed to the new one, so in-flight queries and
  mutations finish on the schema they were admitted under. Operations are
  reference-counted per generation for their whole lifetime (across all subgraph
  hops, and until @defer/@stream streams end). A superseded generation is
  disposed once idle, or force-disposed after drainTimeout; maxConcurrentGenerations
  caps how many generations may overlap. Subscriptions are not pinned: one on a
  superseded generation ends with SCHEMA_RELOAD when that generation is
  disposed — immediately on reload when nothing is draining, otherwise once the
  last in-flight operation finishes (at the latest after drainTimeout) — and
  the client then reconnects against the new schema. Disabled by default.

  export const gatewayConfig = defineConfig({
    gracefulSchemaReload: { drainTimeout: 10_000 }, // ms; default off
  });

Patch Changes

• Updated dependencies [65ecf28]:
  • @graphql-mesh/fusion-runtime@1.11.0

@graphql-hive/gateway-testing@9.0.0

Patch Changes

• Updated dependencies [65ecf28]:
  • @graphql-hive/gateway-runtime@2.10.0