apollo-server
apollo-server copied to clipboard
apollographql
'Records ejected from the Engine Reporting signature cache' error
Description
The following error shows up on the production server a lot. Reloading the server seems to make it stop but it always comes back after some time.
This server is processing a high number of unique operations. A total of n records have been ejected from the Engine Reporting signature cache in the past interval. If you see this warning frequently, please open an issue on the Apollo Server repository.
Expected behavior:
No errors.
Actual behavior:
A lot of errors.
Versions
Apollo server: 2.9.11 Node: 12.13.0 Ubuntu: 16.04.6
Runnable reproduction
Unfortunately I can't provide that since this error only shows up on our production server. Understanding what causes this error and whether any action is required would be greatly appreciated.
This tends to happen when clients are sending unique operations on nearly every request at a relatively high rate. This makes it difficult for a variety of in-memory caches that Apollo Server uses.
Is it possible that you're dynamically generating GraphQL queries that are each unique?
For example:
- Each has a randomly generated aliased name for the field.
- Each has any number of randomly added fields.
- Each has a randomly generated operation name.
- Not using
variablesbut instead interpolating variables as string literals.
Do any of these sound possible? It sounds like you have Apollo Studio, so your dashboard may have some insight into the answers to these questions. In that case, you could also consider contacting our customer support for additional assistance!
Setting cache: 'bounded' when initializing our ApolloServer made this error occur less frequently for us. See docs here.
This is a warning that any Apollo Server can be induced to print if a random client on the internet sends it enough distinct queries. I've never been particularly excited about it and it probably should be removed.