Closed dcadenas closed 4 months ago
I already merged a PR for this. Now just checking because I think the speed of processing events used to be faster
I suspect that the surge that made me do this change was related to a surge in profiles (see https://stats.nostr.band/#relay_users):
And this is the cause https://ron.stoner.com/I_Gained_1_Million_Followers_in_24_Hours/
This is working but I'm letting it run for a while to see metrics before closing.
A load shedding strategy when we are under backpressure instead of full disconnection as we currently do. Currently when we reconnect we grab 1 hour of data from each relay and that's very hard to change and integrate with the backpressure signal from the consumer. Load shedding is easier to skip that hour surge of events on reconnects and also opens the door to future more dynamic/intelligent adaptations under load.