If your graph's Apollo Studio metrics show a larger number of reported operations than you expect, a few different factors might contribute to this discrepancy:
Cached queries, subscriptions, and erroneous requests are logged for metrics. These requests are included in the reported count, even if they are served from cache or result in errors.
Each query or mutation in a batched request is counted separately.
Custom middleware and network functionality, and modifications to metrics instrumentation can effect the number of requests sent.
To help isolate an underlying cause, check whether your metrics discrepancy is limited to a particular variant of your graph. If it is, start by investigating any differences between that variant and your other variants.
Improving Apollo Gateway performance Load testing a federated GraphQL API
Please sign in to leave a comment.