- Chuck Hardy
- Hari Mohanraj
- Graham Phakos
- ≈ 10K exceptions per week (not all showing in tools)
- ? exceptions that might be issues we can action
- Reported exceptions !== actual issue (untrusted, unused tools)
- Ignoring of exception reports over time
- Decreased reaction time
- Users reported issues before we noticed our systems did
- Contributors did not notice negative impact
- Developer happiness reduced
- We look incompetent (I feel)
- < 100 exceptions per week (all showing in tools)
- < 5 exceptions that might be issues we can action (will reduce now that we have visibility)
Also...
- Additional reporting added (Honeybadger, NewRelic, Librato)
- Code Climate GPA increased
- Test coverage increased
- Technical debt reduced
- Reported exceptions === actual issue (trusted, use tools)
- Exception reports are no longer ignored
- Increased reaction time
- Users report issues when we are going out with the fix
- Contributors notice any negative impact immediately
- Developer are happier
- We look competent
- We are in a far better place now than before
- We need to keep working on fixing common exceptions
- Don't get into this situation again
- Exceptions should be rare, don't ignore them, ever!
- If you raise an exception on purpose, ensure it matches what actually is happening and give as much detail as possible. (Sell the Why)
- Take responsibility for issues you cause. Don't rely on Foundry to fix them down the line
- Don't only test (unit or functional) the happy path
- Keep an eye out for possible issues for a few days after your release
- If you think you might have caused a reported issue, say something
- As a reviewer be sure to look out for these points
Your objective is to release to Production without negative impact or additional technical debt. We are now in a situation where the exception reporting tools will help achieve this objective.
