Facebook’s outage caused significantly increased load on DNS resolvers, among other effects. Cloudflare also published this followup article with more findings.

Celso Martinho and Sabina Zejnilovic — Cloudflare

Shell (the oil company) reduced accidents by 84% by teaching roughnecks to cry. Listen to this podcast (or check it out in article form to find out how. Can we apply this to SRE?

Alix Spiegel and Hanna Rosin — NPR’s Invisibilia

Don’t have time to read Google’s entire report? Here are the highlights.

Quentin Rousseau — Rootly

I really like how open Facebook engineering has been about what went wrong on Monday. This article is an update on their initial post.

Santosh Janardhan — Facebook

Want to learn about BGP? Ride along as Julia Evans learns. I especially like how she whipped out strace to figure out how traceroute was determining ASNs.

Julia Evans

The Verica Open Incident Database is an exciting new project that seeks to create a catalog of public incident postings. Click through to check out the VOID and read the inaugural paper with initial findings. I’m really excited to see what this project brings!

Courtney Nash — Verica

Printing versus setting a date — they’re only separated by a typo. Perhaps something similar happened with Facebook’s outage.

rachelbythebay

Adopting a microservice architecture can strain your SRE. This article highlights an oft-missed section of the SRE book about scaling SRE.

Tyler Treat

Despite carefully testing how they would handle this week’s expiration of the root CA that cross-signed Let’s Encrypt’s CA certificate, they had an outage. The reason? Poor behavior in OpenSSL. See the next article for a deeper explanation of what went wrong with OpenSSL.

Oren Eini — RavenDB

This article explains why some versions of OpenSSL are unable to validate certificates issued by Let’s Encrypt now, even though the certificates should be considered valid.

Ryan Sleevi

This says it all:

It turns out that the path to safety isn’t increased complexity.

Matt Asay — TechRepublic

The thrust of this article is that reliability applies to and should matter to the entire company, not just engineering. I really like the term “pitchfork alerting”.

Robert Ross — FireHydrant

Lesson learned: always make your application server’s timeout longer than your reverse proxy’s.

Ivan Velichko

Who deploys the deploy tool? The deploy tool, obviously — unless it’s down.

Lorin Hochstein

Their approach: group tables into “schema domains”, make sure that queries don’t span schema domains, and then move a schema domain to its own separate database cluster.

Thomas Maurer — GitHub

Groot is about helping figure out what’s wrong during an incident, not about analyzing an incident after the fact. I totally get why they need this tool, since they have over 5000 microservices!

Hanzhang Wang — eBay

SRE is a broad, overarching responsibility that needs a multitude of role considerations to pull off properly.

Ash P — Cruform

Here are some things that make SREs a unique breed in software work:

The one about Scrum caught my eye, and I followed the links through to the Stack Overflow post about SRE and Scrum.

Ash P — Cruform

An in-depth explainer on the Linux page cache, full of details and experiments.

Viacheslav Biriukov

There’s some great advice in this reddit thread… and maybe some tongue-in-cheek advice too.

Take production down the first day they give access — then it’s nothing but up from there!

Various — reddit

Using two real-world case studies, this article explains how developer self-service can go wrong, and then discusses how to avoid these pitfalls.

Kaspar von Grünberg — humanitec

What a great idea! I found it especially interesting that only 34% of SRE job postings mention defining SLIs/SLOs/error budgets.

Pruthvi — Spike.sh

For the first time, we’ve created the State of Digital Operations Report which is based on PagerDuty platform data.
[…]
we will walk through some of these findings and share 10 questions teams can ask themselves to improve their incident response.

Hannah Culver — PagerDuty

Incident response so often gets mired in assumptions that need to be re-evaluated. This article uses an incident as a case study

Lawrence Jones — incident.io.

This one lays out clear definitions of SRE and DevOps and compares and contrasts them.

Mateus Gurgel — Rootly

This week, Saleforce released Merlion, a Python library for time series machine learning and anomaly detection. Linked is an in-depth research paper on Merlin, explaining its theory of operation and experimental results.

Bhatnagar et al. — Salesforce

Faced with a difficult hiring market for SREs, they embarked on a well-designed, carefully thought out program to hire and train entry-level folks as SREs — and it worked!

Thomas Betts — InfoQ

No matter how good your tooling is, how experienced you are, or how much you’ve prepared, incidents can still be hard.

Five people share about what they find hardest during incident response.

Chris Evans — incident.io

This one has a lot of ideas about how to guide developers toward full ownership of their services in production.

Ambassador

In this post, I will cover the following modes of system resilience:

• Adaptive Response
• Superior Monitoring
• Coordinated Resilience
• Heterogenous Systems
• Dynamic Repositioning
• Requisite Availability

Ash P — Cruform

Root cause of success: unpatched security vulnerability

TMW a security vulnerability allows you to break into your infrastructure, averting disaster during an incident.

Lorin Hochstein, with incident story by Eric Dobbs

A migration didn’t go as planned, and customer traffic lost its way.

Heroku

I’m a big believer in human-in-the-loop automation. My favorite part of this article was this:

A further problem is that full automation — which aims to take the human out of the picture — requires a complete, nuanced understanding of a system and all potential outcomes, paradoxically resulting in heightened system complexity.

Tina Huang — Transposit

Lots of details about how Slack does incident response in this one.

Stephen Whitworth — incident.io

This list also gives an interesting insight into the way this company does SRE.

Mayank Gupta and Merlyn Shelley — Squadcast

Oh BGP, you rascally little routing protocol.

Alessandro Improta and Luca Sani — Catchpoint

A comprehensive definition of SREs and Site Reliability Engineering, including what SREs do and what makes SREs different from other roles.

The article covers various facets of SRE and acknowledges that SREs can perform many roles.

JJ Tang — Rootly

Another really excellent air accident story with lots of great talk about mental models and confirmation bias. The crew saw lots of disparate indications that each didn’t point to anything in particular and each wasn’t a huge problem on its own. That, coupled with confirmation bias, helped them miss what might seem obvious in hindsight.

Mentour Pilot