Hacker News new | ask | show | jobs
by kyrra 3503 days ago
I think this is because most companies do pager duty wrong. I highly recommend the google SRE book[0] (notes here[1], chapter 11 covers oncall/pager). One thing mentioned in this book is compensation for being oncall. At Google we get fairly decent pay compensation for holding the pager, enough where it can incentivize people to be on the rotation.

(I'm a software engineer at google who is oncall at this moment)

[0] http://shop.oreilly.com/product/0636920041528.do

[1] http://danluu.com/google-sre-book/
2 comments
mikhaill 3503 days ago
Thanks for the sources, I'll have to check them out. We've recently spent a considerable amount of time tracking, analyzing and evolving our PagerDuty alerting[0], learnings of which we've shared in a blog post.

[0] https://goshippo.com/blog/evolution-our-pagerduty-playbook-f...

link
user5994461 3503 days ago
> At Google we get fairly decent pay compensation for holding the pager, enough where it can incentivize people to be on the rotation.

Google is a big company. I expect them to 1) Have people in all timezone so that there is no night shift 2) Have many people on rotation so each individual is rarely on shift.

That's not comparable to smaller companies.

link
kyrra 3503 days ago
Just some anecdotal about my team at Google. I definitely do night shifts (we do 1 week long rotations), but with a 30-minute SLA on responding to pages. I've had bad nights where my first page happens at 11pm, then they keep going until 4am. The better part is that we are oncall once a quarter or so (12 people on the rotation).

The thing is, there is basically a waiting list to join the rotation. Compensation is nice for those that are motivated by it. But it also exposes you to a lot of the infrastructure that you normally don't deal with (so it's a great way to learn).

We have a higher-up-the-stack SRE team that does 12-hour shifts so it's not as bad for them. They handle the larger scale issues that are beyond job specific issues (ex: datacenter issues).

I can understand this sucking if you are on a small engineering team where you can't do things like this (I've got friends at companies that employ < 10 developers, I've heard the stories). I guess I wasn't thinking about it for smaller eng teams where the number of people available to support the product isn't there.

link