•1 min read•from InfoQ
Pinterest Engineers Eliminate CPU Zombies to Resolve Production Bottlenecks
Pinterest identified and resolved CPU starvation issues that affected machine learning training jobs on its Kubernetes-based platform, PinCompute. The engineers traced the problem to an unused Amazon ECS agent, which caused memory cgroup leaks. By disabling the agent, they stabilised performance. This case illustrates the importance of understanding system defaults for effective troubleshooting.
By Mark SilvesterWant to read more?
Check out the full article on the original site
Tagged with
#machine learning in spreadsheet applications
#natural language processing for spreadsheets
#generative AI for data analysis
#cloud-based spreadsheet applications
#Excel alternatives for data analysis
#big data performance
#rows.com
#CPU starvation
#machine learning
#Kubernetes
#PinCompute
#production bottlenecks
#Amazon ECS
#memory cgroup leaks
#performance stabilisation
#troubleshooting
#system defaults
#engineering
#unused agents
#resource management