distributed tracing or centralized logging, which one is actually better for debugging

hey everyone, i've been wrestling with how to best instrument our microservices for debugging. we've used centralized logging, like the ELK stack, for a while, and it's good for spotting errors in individual services. but when a request spans multiple services, tracing seems more intuitive for following the flow. what are your thoughts. have you found one approach easier for root-cause analysis in complex distributed systems. i'm curious to hear your experiences and any patterns you've adopted.