1
0
mirror of https://github.com/yldio/copilot.git synced 2024-12-11 04:20:05 +02:00
copilot/spikes/leak/readme.md
2016-11-25 10:34:08 +00:00

786 B

leak

    1. Spawn a bunch of servers:
    • another-fast: a node with a linear memory leak
    • fast: a node with a linear memory leak
    • slow: a node with a memory leak that grows very slowly
    • plain: a node with no memory leak
    1. Spawn an artillery for each node that loads it with a small but constant stream of requests
    1. Spawn Prometheus that watches the cpu/memory of each node

Then, locally we start the same server and we can see the different instances and an aggregate of the metrics for each job.

usage

λ docker-compose up
λ node .

Go to http://127.0.0.1:8000/ and see the result. The Prometheus is also listening at http://127.0.0.1:9090/

example