Historical content: I wrote this article over 8 years ago on 2008-11-26. You are viewing an archived post from my old WordPress blog. The archive contains over 1,600 articles written over a ten year period. The formatting and contents of the posts may not display perfectly.
The usual trace/error/warning/info advice is totally useless in a large scale distributed system. Instead, you need to log everything all the time so you can solve problems that have already happened across a potentially huge range of servers.