SRE Weekly Issue #349

A message from our sponsor, Rootly:

Manage incidents directly from Slack with Rootly 🚒.

Rootly automates manual tasks like creating an incident channel, Jira ticket and Zoom rooms, inviting responders, creating statuspage updates, postmortem timelines and more. Want to see why companies like Canva and Grammarly love us?:


This is a new spin on dependency-based attacks that has interesting implications for reliability.

  Christoph Treude — The Conversation

In a counterpoint to the articles I linked to last week, this engineer expects that the Twitter infrastructure will keep trucking on for awhile due to automation and redundancy.

  Rory Bathgate — ITPro

When folks use blaming language, bring up counterfactuals, or exhibit other suboptimal behaviors in a retrospective, what’s a good way to respond, and what doesn’t work as well?

  Fred Hebert

Cloudflare uses a novel approach to make the most out of a limited number of IPv4 addresses for outgoing traffic: “soft-unicast”.

  Marek Majkowski — Cloudflare

After attending my first retrospective at Honeycomb, I wrote this article about how they establish expectations and shared context at the start of the meeting.

  Lex Neva — Honeycomb

Pilot incapacitation, an argument, and a broader rift between cohorts of pilots were just some of the many contributing factors in this air accident. In response to this 1972 accident, the UK mandated cockpit voice recorders on all commercial flights.

  Admiral Cloudberg


Categorized as SRE