Refine the urgencies defined for GraphQL queries
In #331697 (closed) we've added apdex metrics for GraphQL. These currently all use a default target duration of 1 second coming from the default urgencies we've defined for requests
Now, we should revisit these urgencies and make them configurable by the groups that own them. Looking at the overall distribution, most queries perform way faster than requests: https://log.gprd.gitlab.net/goto/c4c06910-6d78-11ec-bf16-136dbae1ae92
My proposal would be to have the following
Urgency | Target duration | Cumulative percentage of traffic | Note |
---|---|---|---|
High | 0.25s | 99.463% | |
default | 0.5s | 99.594% | Default for known requests from our frontend |
medium | 1s | 99.69% | |
low | 5s | 99.991% | Default for unknown requests from external clients |
To customize an urgency for a known requests, we could allow specifying this using a magic comment in the .graphql
query file.
Edited by Bob Van Landuyt