The #535 PR was a first step towards working with real traffic patterns. The idea is that if we capture request timings and contents from the devnet (and beyond) and then have a way of replaying it exactly how it happened, it would make debugging load issues much easier.
So further actions as I see it are twofold:
- Have a more compact way of serializing requests than what we have in logs, and maybe have that mirrored via a sink at a separate endpoint
- Introduce a replay tool that would be able to take those logs and fire them at our stack with the same timings as how it happened in real time on devnet, so that it can be reproduced locally
The #535 PR was a first step towards working with real traffic patterns. The idea is that if we capture request timings and contents from the devnet (and beyond) and then have a way of replaying it exactly how it happened, it would make debugging load issues much easier.
So further actions as I see it are twofold: