#Prevent Extra Workers from appearing

11 messages · Page 1 of 1 (latest)

obsidian vector
#

Many times extra workers are spawned for multiple hours even though there is no need for them as the load is easily kept up by the normal workers. How can i prevent these from appearing?

i already set max workers but it does not help. this costs so much money that i am thinking about switching provider.

west birch
#

Extra workers that aren't running are free

#

So they don't cost you anything if they're idle

#

They are for extra measure to handle throttling, if many workers are throttled you'd still have the normal amount or more workers than without them

jolly heath
#

Extra workers never run at all, they are only there to help with throttling.

#

RunPod probably needs to clarify the cost somewhere because seems a lot of people somehow think they are charged for max and extra workers like with active workers.

west birch
#

oh i thought they also run hahah

#

yeah

jolly heath
#

Nah if you have 3 max workers + 2 extra workers, only the 3 max workers will handle requests and the queue will build up and the 2 extra will sit idle if all 3 max workers are handling requests. You cannot have more workers running than your max workers.

mild tartan
#

Extra workers are pre-provisioned but do not run unless necessary. They are designed to handle spikes in load by being available to start quickly if all max workers are busy. If you have set a limit of 3 max workers and have 2 extra workers:

•    Normal Operation: Only the 3 max workers handle requests.
•    During Throttling: If the load exceeds the capacity of the 3 max workers and they are all handling requests, the extra workers can be activated to manage the additional load.

This setup ensures that the system can handle sudden increases in demand without immediate throttling, improving responsiveness and stability.

Extra workers do not incur costs when they are idle. You are only charged for the workers that are actively handling requests. This allows you to have a buffer for handling spikes without incurring extra costs when the demand is low.

@west birch @jolly heath is above explanation correct?

west birch