API / Apollo Gateway Interview questions
Explain the lifecycle of a GraphQL request through Apollo Gateway?
A request passing through a federated gateway goes through a fairly fixed sequence of stages before a response comes back to the client.
flowchart TD
A["Client sends GraphQL query"] --> B["Gateway parses and validates against the supergraph schema"]
B --> C{"Query plan already cached?"}
C -->|Yes| E["Reuse cached query plan"]
C -->|No| D["Query planner builds a new fetch plan"]
D --> E
E --> F["Execute fetch plan: sequential and/or parallel subgraph calls"]
F --> G["Entity representations passed between dependent fetch steps"]
G --> H["Responses from all subgraphs stitched into one result tree"]
H --> I["Errors, if any, merged into the errors array with corrected paths"]
I --> J["Final response returned to client"]
Two things are worth calling out: query planning is decoupled from execution — the plan only depends on the query's shape, so identical queries reuse a cached plan — and stitching happens after all subgraph calls return, matching the shape of the response back to exactly what the client originally asked for, regardless of how many services contributed to it.
According to the request lifecycle, what determines whether a cached query plan can be reused?
At what stage are subgraph responses combined into the final result tree?
More Related questions...
