Looking forward to @drensin talking about how SRE relates to DevOps, on All Day DevOps alldaydevops.com/nowlive
I’ve seen Dave speak twice at SREcon Americas and he’s great.
“Reliability is the most important feature of any system.” @drensin #alldaydevops
Users won’t care if you say everything looks good on your end, if they’re having a bad day. The only perspective that matters for measuring reliability is your users’ perspective. @drensin #alldaydevops
Users won’t notice any reliability you have, over the least reliable thing between them and your system. @drensin #alldaydevops
To add an extra 9 of reliability is an order of magnitude increase in cost, at least 10x. @drensin #alldaydevops
SLAs are a business discussion, not an engineering discussion. @drensin #alldaydevops
The reason we go oncall is to learn what needs to be automated. @drensin #alldaydevops
You should think of SRE as a highly opinionated implementation of DevOps (class SRE inherits DevOps). @drensin #alldaydevops
Nice @lizthegrey shoutout in @drensin’s talk :) Liz is awesome and has done a ton to evangelize SRE.
Enterprises love SRE because they understand TCO and ROI. @drensin #alldaydevops
Error budgets and SLOs prevent “intuition fatigue.” @drensin (I love that term! And have definitely experienced it.) #alldaydevops
SRE principles incentivize you to reduce complexity. And simpler systems are easier to reason about. @drensin #alldaydevops
Ahh this talk is so good. I really recommend that people interested in SRE track down the recording when it’s posted. Especially if you are newer to the concepts. @drensin explains them so well. #alldaydevops
A pacemaker needs 3 1/2 nines of reliability. Good rule of thumb to measure against. @drensin #alldaydevops
Start with one application. So that first. @drensin (This is a good idea for many things, like implementing config management or other automation. Pick one thing and make it work.) #alldaydevops
(Once you have a success, you can use that to evangelize internally. People often need to see something work before they will buy into it.) #alldaydevops
Dave linked to this spreadsheet for calculating risk. docs.google.com/spreadsheets/d… @drensin #alldaydevops
More logging and measurement is probably better. More alerting is probably not. Only alert on the symptoms of your users’ pain. @drensin #alldaydevops
Alert on SLO violations, or if you’re burning your error budget too fast and the SLO is in jeopardy. Alert on things your users would want to know about. @drensin #alldaydevops
Dave is talking a lot about blameless culture to wrap up. You won’t learn from mistakes if people don’t feel safe being honest. @drensin #alldaydevops
This is so important. I’ve worked in shops where people were worried about getting fired if they make a mistake. There are so many negative results from a culture like that. #alldaydevops
• • •
Missing some Tweet in this thread? You can try to
force a refresh
