Career-long sysadmin/SRE/SRE Team Lead, here. I've worked at large shops (10-30 million end users) and some shops where 99.98% is the SLA to prevent millions of dollars of losses in supply chain.

First of all, I appreciated this diary because Anna took the task with a positive attitude and as a learning experience. Thanks for writing this. To see an old problem through new eyes is inspiring.

I have numerous, "hot-take" criticisms of your current organization's practices, but I'm not sure I have all the context yet. The one suggestion I will make is: if you're not already using it - clone https://github.com/pagerduty/incident-response-docs/ and modify it to meet your organization's needs. Then, have it blessed as policy by management and train SREs and Devs on it.

To the other comments: I see there's a lot of people here who say they'd never do the SRE job, or return to doing it. I'm not discounting your fear or feelings of burnout. Been there. But, hear me out:

DevOps is not just about CI/CD pipelines and monitoring and Pagerduty. It's about having a culture where developers don't throw operational or security poop over a wall of confusion at sysadmin types as well as at their peers. This kind of organizational dysfunction can be devastating to a business.

DevOps at it's best is about about empathy. One of the best places I ever worked was filled with developers who had true empathy. They realized that an error or omission in their work could would wake up their Ops team at stupid o'clock in the morning, repeatedly - leading to all the things that drive SRE's and on-call folks literally insane. They practised strict TDD.

These developers volunteered to be second-tier on call after the ops team did triage, out of the kindness of their hearts for their coworkers. Management also led a culture of defending time to find permanent solutions to drive measured improvements in SLI.

SRE isn't about waking up at stupid o'clock every night to press buttons. It's about having a culture of driving permanent fixes and compensating by using cost-effective and appropriate cloud architectures. It's also about leading the working agreements with engineering teams to do blameless post-incident retros together and making the work bring your teams closer instead of pushing them apart.

I can't help but take away that a lot of you feel like On-Call heroics are what SRE is about. It's more difficult than that, but also less stressful, and simultaneously more rewarding when you get it right.