Our current logging system needs improvement for a couple reasons:
1. When things fail, it's hard to pinpoint the exact location of failure. Very often the errors bubble up (e.g. JAVA exceptions) and by the time it is noticed, it's hard to tell exactly where the failure occurs. We need some way to make it so that the exact errors can be easily identified (e.g. reserve some unique string such as "ALERT")
2. The logging system is supposed to be layered with one layer encompassing all messages from another layer. However, at some point, we have lost that layered design and many messages would occur in system or error logs but not in the debug log.
3. basic misconfiguration issues are not spotted right from the start (system down, etc.)