increase api nodes cpu utilization by adding more unicorn workers
When investigating api latency issue we determined that the queue is growing because there are no free unicorn workers on api nodes.
One approach to fixing that is adding more api nodes. Another is adding more unicorn workers per api node.
The CPU usage on api nodes is relatively low:
disk IO is also relatively low:
This issue was created to discuss the feasibility of adding more unicorn workers on api nodes to utilize the otherwise idle resources.
Edited by 🤖 GitLab Bot 🤖