What error reporting method do you use for your applications?

[Kyjan]

Greetings!

So, I'm curious to know what type of error reporting your applications use in your work environment. Obviously, there may be only so much that you can say; however, a high-level view would be helpful. For example, do you use a middle-tier message queuing application to ensure error delivery from your application? Do you send emails to a monitored email box every time something happens? Do you load all of the errors into the database?

What would you change if you could to make it better?

Thanks,

Kyjan

[techgnome]

In the last place where I worked, we used two notifications... the first and primary notification is a user-friendly, something's gone wrong error, and a second, much more detailed, and technical - to include the call stack if possible - message that gets logged to the system's event viewer in the database.

-tg

[Hack]

the first and primary notification is a user-friendly, something's gone wrong error

We do this as well, but also log the error to an error log database table. The info stored is the error number, description, the name of the application in which the error occured, place in the application where the error occured, name of the logged in user and time the error occured.

[Merrion]

Depends on the type of application.

For command line type things we use a job scheduler (Autosys) that has built in error logging and reporting.

For front end apps we have TraceListener derived classes that can log errors to the event log and also email them.

[Kasracer]

We do this as well, but also log the error to an error log database table. The info stored is the error number, description, the name of the application in which the error occured, place in the application where the error occured, name of the logged in user and time the error occured.

What happens if the error is related to your database?

[techgnome]

The error still gets logged.... why wouldn't it?

-tg

[Kyjan]

So do you use a custom front-end application to view the information in the database?

Also, I haven't noticed any use of something like Microsoft Message Queuing. How do you ensure that the errors actually make it to the logs? What happens if the system hits a timeout?

Kyjan

[techgnome]

well, first, there shouldn't be a reason for time outs... and secondly... it still gets logged. It goes into a local cache that gets updated to the database later.

-tg

[Hack]

What happens if the error is related to your database?

Sure. A database error would cause an error in the front end code, which, in turn, would get logged.

So do you use a custom front-end application to view the information in the database?

For the longest time we were just opening Enterprise Manager and doing a Open Table/Return All Rows but that got a bit tiring so I did finally build and error log screen for us.

[Kyjan]

Thank you all for your responses thus far. It is really helping me to get a grasp on how I want to report my errors. Now, if I may, let me branch out on a related topic.

How do you handle information that you want logged? Do you log it in the same manner (to a database)?

well, first, there shouldn't be a reason for time outs... and secondly... it still gets logged. It goes into a local cache that gets updated to the database later.

What if you're connecting across the network to another server located somewhere else? Wouldn't there would be the potential for timeouts? Also, does your local cache work on a FIFO basis? Or do you do priority based logging?

[techgnome]

In short, the architecture is such that latency isn't a problem in connections - all connections remain within the same network.... but if it still becomes a problem, odds are, that means NO ONE is online.

To answer the other question, FIFO I guess.... it's part of the architecture... and as far as I know, it's a batch operation.

-tg

[Hack]

What if you're connecting across the network to another server located somewhere else? Wouldn't there would be the potential for timeouts?

As long as it is on the same network there shouldn't be a problem.

[mendhak]

I wouldn't apply a single error reporting mechanism to all applications. Different applications have different needs in terms of what the user should or shouldn't see when an exception is thrown. Minor applications may be happy by simply logging to a text file, other applications may need the event logged to the event viewer and the email sent out.

[Hack]

Minor applications may not even have error logging (error trapping yes)

[techgnome]

our logging was dictated by Sabanes/Oxley .... we had to record EVERYTHING... ugh...

-tg