fix(federation/query-planning): Prevent error for satisfiable @shareable mutation fields

[sachindshinde]

This PR fixes a bug where query planning may unexpectedly error while planning an operation with a satisfiable @shareable mutation field.

Specifically, this PR:

1. Updates query planning on mutation fields to only consider one initial subgraph at a time for a traversal, and returns the plan with lowest cost.
   • Previously, this was mixing up options that started in different subgraphs in the same traversal, which would sometimes generate query planning errors if those differing options were chosen. There were a few ways to fix this, but the least invasive/simplest was just to do multiple traversals, each with a limited initial subgraph.
   • Non-local selection estimation was also appropriately updated to account for this (if it weren't, it would systemically overestimate the number of non-local selections).
2. Updates a particular query planning optimization to avoid discarding an option if another option with less subgraph jumps/cost is found, if that option starts in a different subgraph and the option is for a mutation operation.
3. Updates query planning to ignore @key edges at top-level (previously only root type resolution edges were ignored).

This PR is the equivalent of this JS repo PR.

Unrelatedly, this PR also fixes an error messaging issue when QP unexpectedly doesn't find a plan (this shouldn't happen, but if it did, it was previously using the wrong federation error enum and message).

[apollo-librarian]

✅ Docs preview ready

The preview is ready to be viewed. View the preview

File Changes

3 new, 10 changed, 4 removed

+ graphos/routing/(latest)/observability/router-telemetry-otel/apm-guides/datadog/observing-and-monitoring/dashboard-template.mdx
+ graphos/routing/(latest)/self-hosted/managed-hosting/railway.mdx
+ graphos/routing/(latest)/self-hosted/managed-hosting/render.mdx
* graphos/routing/(latest)/customization/rhai/index.mdx
* graphos/routing/(latest)/customization/coprocessor.mdx
* graphos/routing/(latest)/observability/graphos/graphos-reporting.mdx
* graphos/routing/(latest)/observability/router-telemetry-otel/apm-guides/datadog/connecting-to-datadog/datadog-agent/datadog-agent-metrics.mdx
* graphos/routing/(latest)/observability/router-telemetry-otel/apm-guides/datadog/connecting-to-datadog/datadog-agent/datadog-agent-traces.mdx
* graphos/routing/(latest)/observability/router-telemetry-otel/apm-guides/prometheus/otel-traces-to-prometheus.mdx
* graphos/routing/(latest)/observability/router-telemetry-otel/enabling-telemetry/selectors.mdx
* graphos/routing/(latest)/observability/router-telemetry-otel/enabling-telemetry/spans.mdx
* graphos/routing/(latest)/self-hosted/containerization/docker.mdx
* graphos/routing/(latest)/_sidebar.yaml
- graphos/routing/(latest)/observability/router-telemetry-otel/apm-guides/datadog/index.mdx
- graphos/routing/(latest)/observability/router-telemetry-otel/apm-guides/datadog/router-instrumentation.mdx
- graphos/routing/(latest)/observability/router-telemetry-otel/apm-guides/datadog/connecting-to-datadog/index.mdx
- graphos/routing/(latest)/observability/router-telemetry-otel/apm-guides/datadog/connecting-to-datadog/otel-collector.mdx

Build ID: 8be98f32a1d4924fb5228093
Build Logs: View logs

URL: https://www.apollographql.com/docs/deploy-preview/8be98f32a1d4924fb5228093

[dariuszkuc]

Corresponding JS fix

[sachindshinde]

@dariuszkuc
That PR has been split up, and this PR corresponds more closely to apollographql/federation#3304 now.

[duckki]

Looks great! The fix makes sense and I confirmed the test is now fixed.

[sachindshinde]

After further testing, there were two more changes needed:

1. Top-level query planning was ignoring root type resolution edges, but not @key edges; this has been fixed now.
   • This normally wouldn't be an issue as @key on root operation types is inefficient compared to root type resolution edges, meaning query planning wouldn't normally choose such options. However, for supergraph schemas with unsatisfiable mutation operations, this could accidentally generate a plan instead of an error (when it should be erroring).
2. Non-local selection estimation needed to be updated to accommodate initial subgraph constraints, otherwise it would systematically overestimate the number of non-local selections.
   • If not fixed, this leads to mutation query planning unexpectedly experiencing a QueryPlanComplexityExceeded error when there are many subgraphs with mutation types.

[duckki]

The follow-up update also looks good. Nice work!