Post by @jorijn
#kubernetes HPAs working in harmony keeping the #tootcommunity people happy.
What are we looking at?
It tries to keep the average number of pending jobs in each queue below 500, if it goes over, it starts additional processes that pitch in with the workload.
Web processes are managed so that no individual process handles more than 5 req/sec. More get added if it goes over.
If the amount of work is decreasing, capacity is returned to the cloud so we don't pay for it. 1/2
#devops #mastoadmin
What does this mean for costs?
During peak times it costs $0.78/hour to run #tootcommunity, but only $0.26/hour during the night, thanks to being able to dynamically purchase additional servers and return them when they're no longer needed. 2/2