kuberenetes cluster autoscaling, metrics and strategies

hey everyone im trying to nail down the best way to configure autoscaling for our kubernetes clusters weve seen spikes in traffic that arent always perfectly aligned with cpu or memory usage and im wondering what other metrics or approaches you all use to trigger scale up or down events are you primarily relying on kubernetes metrics or are you integrating with external monitoring tools also how do you handle the scaledown part to avoid thrashing or killing pods that are still busy curious to hear your experiences