Closed Jared-Sprague closed 6 years ago
investigation notes: Using the version of the server in June: commit 1fe6df03230d57c1c41be73516a73bcc803f9636 Clients, CPU 2, 6 4, 7.2 6, 8.5 8, 9.5 10, 12 12, 14 14, 25 16, 30 18, 70 20, 80 22, 100
Using latest version of master at tag 2.0.0-4: Clients, CPU 0, 3 2, 6 4, 7 6, 9 8, 11 10, 11 12, 14 14, 20 16, 40 18, 70 20, 90 22, 97
Interesting, in the latest version when I kill all the test spheres the cpu stays high, because their clients are still connected, they are just not sending position updates. suggesting that perhaps most of the cpu load is from broadcasting, not receiving:
I noticed that there is a leak somewhere in socket_uuid_map:
Traced this down to a bug in node-game loop in the latest version: https://github.com/tangmi/node-gameloop/issues/4
Downgrading to 0.1.2 of node-gameloop fixes the problem
Yep definitely fixed:
Not a lot changed server side, except for the Bot paths. If I had to guess I'd say this is related to bot paths.