How you tackle a defect in your code, specially a server side code? Are you remotely debugging the code at first step? Yes it is convenient to do so and most of the time it save you from the defect. The real question is, can it save you from defects all the time? Imagine you have written a code which is highly dependable on external factors such as data base connection, server level configurations and fortunately all goes well in development environment as well as QA environment. Assume you suddenly got a complaint from the customers that your functionality is not working on production as they expected. How you tackle where it went wrong, are you still willing to perform a remote debug procedure on the production where we need to maintain high availability? Who’s gone to save you now? Don’t worry logger is here. Most of the time it save us from incidents (not all incidents are bugs)My ABC of defect tackling on a server side code.
- Try to find any unexpected event on server log at the time the issue was raised.
- Identify the last known good server log and compare rest with code.
- If neither of above is able to solve the incident, remotely debug the application.
So it is very important to maintain a good server log. My opinion is we should be able to tackle any issue without remotely debug the code.Impact on application performance by java logging.
Logging has big influence on application performance since logger has to perform IO operations, string concatenations each time we ask logger to log a message so most of the developers avoid them but it is not advisable to avoid all of them. What we can do is to identify what are needed to be log and what are not. Keep in mind both less and excess logging is not preferable.Determine the correct type of before log.
I’m using log4j as my logging provider but SL4j is grooming as the standard logging API. In log4j there are four types. It is really important to place log statement on correct category.DEBUG : The lowest restricted logging type , ideally we should include all the information we need to debug the application but keep in mind DO NOT enable this level on production because it heavily impact on application performance.
INFO : More restrictive than DEBUG and only for use to log informative events happens during the execution. Examples :
WARN : More restricted than INFO and used to log warning and alerts. Examples : ERROR : Most restricted and used to log exceptions. We need to take those logs as more serious and we must log more details to find out what is wrong.
Important facts you should consider while logging.
Minimize string concatenations whenever you can.
String concatenation is a costly operation we should minimize concatenations, even we are using INFO as our debug level so logger is never log anything mark as DEBUG but in JVM it is executed . As I mentioned above DEBUG should contain all the details we should use to tackle the issue so most probably there will be many string concatenations. Example : Solution 1 : Check debug level when there are too many string concatenations . Solution 2 : Use parametrized logger framework such as SL4J.
Minimize possibility of throwing exception while logging.
Example : There are two possibility of getting NPE in above log statement. I’ve seen scenarios where actual exceptions are hidden by the exceptions thrown by the logger statements so try to avoid them because loggers are here to help us not to worst the situation.
- Make sure logger is not null before log. Sometimes we are accidently using inherited logger instance which may not be initialized.
- Make sure “person” object never be null in above statement.
Use descriptive log statement.
Is there any use of follow log statement?
- Never log sensitive data such as passwords, credit card numbers.
- Avoid spelling mistakes and try to be grammatically correct all the time.