-
Notifications
You must be signed in to change notification settings - Fork 141
Description
While adding missing metrics for EL (and logs for EL/CL) I've noticed that for some routes (AttestationData fetching, for example) we seem to have quite high request rate per second (10+),
we might want investigate this further to see if some of the routes can use caching (similar to the kind we do for AttestationData CL calls), especially if request takes > ~10-100ms (but even if it takes less - the sheer number of them can be problematic as well).
We can find all the relevant CL/EL requests by querying for the following log-lines (also note, some of these might be actually throttled to 1% or 10% of their actual occurrence simply to keep the logs growth at bay):
CL single-client request done
CL multi-client request done
EL single-client request done
Also note, these logs (and corresponding metrics) ^ are supposed to track only "real requests made", (I guess "real" here means a call into 3rd-party client library, not necessarily an HTTP/RPC call, since we don't have any control of what that code in there does - they might be doing their own caching, as seems to be the case for CL "Domain" route)
We might also consider pre-fetching some data, eg. "Domain" call seems like a good candidate for that.