The August meetup will be held at Facebook's new Seattle location on 300 8th Street in Seattle. Please bring a government-issued photo ID to check in with security. When you arrive at 300 8th Ave N, enter the building (Arbor Blocks 300), check in at the security desk, and a host will bring you over to the meetup. Looking forward to getting everyone back together after the holiday break!
Speakers for the evening:
Danyel Fisher, Honeycomb.io -- "EVERYTHING IS BROKEN -- but how broken is it?"
When do you drop everything to fix an error, and when can you leave it for tomorrow? This can be not just a technical question, but one that crosses from engineering priorities, through the C-level suite, and through marketing. The idea of keeping a “reliable” system can be a point of negotiation and contention. Any service has a series of small continuous hiccups and alerts — how do we keep that from turning into burnout?
In mid-July, Honeycomb had twelve minutes of outage for our data ingest process. Is that a lot, or a little? Our incident review showed no less than five independent contributing causes that led up to that moment. Does that mean that we were doing great — or that everything was terrible? A Service Level Objective is meant to help describe how bad things are allowed to be. In this talk, we’ll discuss how establishing a Service Level Objective can provide a way to prioritize errors, and let you sleep better through the night.
Ravi Lachhman, Harness -- "Human no, machine yes – Using ML in Continuous Delivery to Give the Power to the Developer"
The usual method of delivering software into production consists of developers creating the artifact, then handing it off to a team of DevOps engineers who sweat bullets as they write scripts and carefully deploy code into production. This method is time-consuming, error-prone, and is no longer suitable for the accelerated pace of business.
In this talk, Harness discusses the need to combine the practice of Continuous Delivery with Machine Learning to give power directly to the developer team -- including actually how to do it.
We will discuss:
- How to create a failure strategy around product deployments, giving the business confidence that "push-button deployments" will not result in performance problems
- How both response times and custom events can be teased out of existing monitoring and logging tools using the correct algorithms
- How to create a dashboard that gives engineering leaders visibility into every deployment
*** Interested in hosting or speaking at a future meetup? Reach out to the organizers!