Comment by site-packages1 | Hacker News Reader

Comment by site-packages1 | Hacker News Reader

Parent Full thread

I don't do this anymore. Very draining to be an SRE. I remember one Christmas I worked until 2am because of SRE pages on software that, in retrospect, didn't need to even be up at the time. Google used to give us SRE bomber jackets, and while those were super cool in a nerdy way (I never got one because I wasn't an SRE there), it was clear the org had to really go out there with incentives to get people to sign up for SRE. I love the practice of reliability and building fault tolerant systems, but it's just so draining / energizing at the time, then you realize all your efforts (or at least mine and colleagues' at the large companies for which we worked) went into keeping stupid, unimpactful things from falling over briefly. Unless you're SRE for a real, not startup-tongue-in-cheek "mission critical" system it is absolutely not worth it. In my very cynical opinion, I'd need to be an SRE keeping a NASA rocket working, or keeping the ISS up, or working on some SDV system, before I would consider a system to be really mission critical.

It’s only mission critical if someone is going to possibly die (as you mentioned life critical systems). Otherwise, you’re just burning up quality of life for someone else’s profits and arguably weak comp (compared to the family and health harm of SRE roles, and more broadly speaking, on call in general with absurdly short response and RTO SLAs). This isn’t doctor pay except perhaps at FAANG.

(general observations over a longish career)

Well said, that's largely what I was getting at when discussing NASA and self driving vehicle type of applications.

Well the alarm would have been 20 minutes after when it actually happened on Mars.. and any solution won't be there before a same amount of time.

Especially for their helicopter, rover at least stays on the ground throughout all this.

Being on-call I have developed insomnia to the point I'm considering taking sleeping medication nightly...except, of course, the week-nights I'm on call. On-call is just not healthy.

At one point the same applied to our oncall shifts, I'm so glad our managers were able to set apart at least 6 engineer months to tackle oncall issues. Gotten a lot better since then.

I'm curious what sort of response SLO your team had. My understanding was that the 5m and 30m response tiers always came with a multi-site "follow-the-sun" rotation.

Yeah, we’re too small to have that kind of rotation, and consequently our SLO’s are business hours only! If the system is actually on fire / down outside that time then we’ll deal with it, but that only happens maybe once a year at most.

Reply on news.ycombinator.com