centralized versus decentralized logging in microservices

hey everyone, i'm wrestling with a common microservices challenge, logging. we're growing, and managing logs across dozens of services is getting hairy. some teams advocate for a centralized logging system (like the ELK stack, or Splunk), while others prefer decentralized logging where each service manages its own logs and tools. what are your experiences. what are the pros and cons you've encountered with each approach regarding maintainability, cost, and debugging effectiveness