Help with odd spike in Apollo Router query_planning time

Router
,
1

Hi,

I have been testing using apollo router in place of grafbase-gateway to route to my subgraphs, but I started running into issues with the query planner of the apollo router.

I have 13 subgraphs running locally, all written in Rust in case that changes anything, and if I query them through apollo router the query planning stage takes 30 seconds to complete.

image
image2442×796 94.3 KB

The same query is instant with grafbase-gateway.

Is there anyway to get more log information from the query planner to see what is happening here?

This is with the latest router release as well 1.59.1

Kind Regards,

Ashley

2

Using the Otel attributes provides the details for time. which is what you have already. There is also the aggregated metrics.

What is interesting is that the busy time of that plan was 0.139ms, so pretty quick. There was just a lot of idle time.

What helps with all issues is actually a link to reproduce the error or at a mimimum the operation and supergraph you used that caused a long planning time.

Were you running other significant load through the Router at the time of operation?