Unpopular opinion on fixing issues:
Investing in DevOps tools to detect incidents and hiring qualified engineers to fix them is not the right way to do it.
Why do I see that as a problem?
I believe that simply detecting issues and hiring qualified people is necessary but not sufficient.
Let’s think about a four-nines SLO – that’s 4.4 minutes a month.
Every issue will break your SLO, as it might take:
- 10-15 minutes on average to get an issue to somebody, and then
- up to an hour to fix it.
So there's no chance you can do anything except dissatisfy people whenever you have an issue.
The only real solution to that is automation.
That’s why I started Shoreline to build incident automation to help people:
- automatically fix issues in production
- grow the number of people who can safely fix things without escalation
For issues that require human judgment, the first person who looks at the issue should be able to resolve it rather than routing it through 6 other people.
And having a system take care of the mundane work is way better for me than dealing with something the 5th or 25th time.
Because, like everyone, I like sleeping at night.
