Observability is the new black.
Just a quick recap of the problem and solution space for serverless applications: how do you observe what is happening inside your highly distributed system of managed services?
Well, there was a bit more to it but that was the theme for this year’s ServerlessDays PDX inaugural event. A quick summary of each talk from Monday:
I can haz serverless? A real-world example of integrating serverless into existing infrastructure
Chase Douglas (https://twitter.com/txase) and Ian Scofield
Great way to kick off the day with a demo of creating more than a hello world serverless application with a little help from Stackery. Unfortunately, the venue didn’t have wifi to use so this was the first of many without the ability to live demo but it was executed as well as can be expected. Stackery gives you the ability to visually observe your cross-cloud infrastructure all at once.
Serverless and DevOps, from the AWS horse’s mouth
Chris Munns (https://twitter.com/chrismunns)
Due to technical difficulties with his laptop, Kelsey gave up the floor to Chris. It was obvious he had done this presentation many times before; great pacing and the slides were a nice compliment.
DevOps is dead, according to Chris. Personally, I think DevOps has finally become what it should have been from the beginning. It is the concept of code ownership instead of handoffs. Serverless forces everyone’s hand in that area because there isn’t much to operate other than your app. Chris says operations will completely go away by 2025.
Going serverless with Google Cloud Functions
Kelsey Hightower (https://twitter.com/kelseyhightower)
“This is what happens when you get rid of all the servers” Kelsey said when he couldn’t get his Google device to work without wifi. His message changed once he got a working system to present on.
Serverless is about one task per app. We’re moving to an environment where we can piece code much like Legos, but at a higher abstraction. Basically we’ll be able to hook SaaS systems together without any coding. That’s Kelsey’s vision of the future.
Observability (and responsibility) for serverless systems
Charity Majors (https://twitter.com/mipsytipsy)
Those who don’t know Charity need to know that she’s been trying to solve distributed systems problems for the better part of a decade. She was instrumental at Parse before and after they were acquired by Facebook, which brought her to co-found an observability company.
If there is one takeaway from what Charity had to say, log as much as you can about everything. A good test is to ask yourself this question: Can I answer new questions about my system without deploying new code? Favor high cardinality with lots of identifiers to debug the unknown unknowns.
Observability for better applications
Erica Windisch (https://twitter.com/ewindisch)
Erica from IOPipe was keeping with the theme of the event explaining that observability was the only way to debug and keep tabs on a distributed serverless system. This was also the second time a speaker referenced the 12 factor app and how serverless enforces it. Can’t be a coincidence.
Enterprise challenges for broader serverless adoption
Eduardo Laureano (https://twitter.com/eduardolaureano)
After lunch we were all in the mood for a love story and Eduardo provided one wholeheartedly. He pointed out a few danger zones for enterprises adopting serverless and how Azure Functions helps resolve them. There was a live demo planned of the Azure SaaS offerings but the lack of wifi demolished yet another speaker’s plans.
Nonetheless, he provided a compelling case for utilizing what Azure has to offer for debugging, long running stateful apps, and distributed tracing. They also provide many slices of infrastructure from an all Azure cloud solution to on-prem on a boat. Yep, on a boat.
What we should all worry about when monitoring serverless applications
Nitzan Shapira (https://twitter.com/nitzanshapira)
Big takeaway from what Nitzan said: serverless time is money. This might be obvious to some but these days, developer time is the most valuable aspect of engineering we have. I was able to speak to Nitzan during lunch and noticed the confidence he has in his product and the solution he is providing. We all need observability into our systems.
During his demo, a couple of key features of Epsagon stood out to me — it provides essentially what is a stacktrace of your system and a cost estimator of all your serverless pieces. Both are important aspects of building these systems to keep in mind. I’ve encountered problems with both over the last couple of years.
Serverless: the missing manual
Will Plusnick (https://twitter.com/WillPlusnick)
During his time, Will pointed out the differences in building serverless applications compared to established technologies. He built a distributed app to provide maps to firefighters of what homes had been burned during a wildfire. Too bad there was no wifi for him to demo it.
Serverless best practices
Chris Anderson (https://twitter.com/jchris) and John Miller
Big takeaway from the folks at FaunaDB: they know databases. Overall, I was most impressed with their deep knowledge of how you should use different databases and the interesting possibilities FaunaDB brings. Unfortunately, most of their serverless points had already been made earlier in the day, but expounding on their product was very compelling.
Lambdas in the real world: serverless development as part of a team
Toby Fee (https://twitter.com/tobyfee)
Serverless development on a team is about compromises. I’ll be honest, I was in the post-lunch lull during Toby’s talk, but one thing I have to say — if you have a chance to see Toby speak, run don’t walk.
Your apps have gone serverless. Has your security?
Tal Melamed (https://twitter.com/_nu11p0inter)
Each time I’m exposed to some moderately deep security demos I am reminded how much I don’t know about security tools and methods. Tal provided a compelling demo showing how easy it can be to allow your serverless system to be compromised if you don’t focus on security.
Principle of Least Privilege people, always.
During the entire day I felt very lucky to be one of the few who were a part of ServerlessDays PDX. The attendee to speaker ratio was much higher than any other event I’ve attended, even a similar event created entirely for my company. I could easily talk with any of the organizers or speakers during breaks.
I also felt like I was a part of something special because the infancy of most of the tools around serverless is at a perfect adoption point. Enough to get the job done, but also enough space to provide your own solution to a serverless problem.