Handling Errors in Distributed Systems with Retries

Now that you have a micro service, you need to rethink how to handle errors, how do you? Would you do a retry? What if the retry fails, what if the service is too slow and so many clients are now doing retries, the load would increase, how do you handle it then?