[cs615asa] [CS615] Meetup Summary and Notes: DevOps for K8S + Operational Excellence in Apr Fools Pranks

[Brian Zawisza]

I also attended the same meetup on March 20th as some other students which
featured "DevOps for Kubernetes" and "Operational Excellence in April
Fools". Here is the link to the event:
https://www.meetup.com/nycdevops/events/246346445/ . Unfortunately the
slides are not yet published, but should be soon!

Why I attended this event:
I have been working with Docker for a while now and have heard Kubernetes
gaining some population this year. I unfortunately missed an internal tech
talk at my company and was hoping that this talk would be a nice
introduction. The first talk caught my interest. The second talk described
how to deploy, prepare, plan, and mitigate risks of a secret high risk
April Fools deployment that effects millions of users. The event abstract
mentioned rollbacks, feature flags, and other techniques to
prevent downtime and ensure a successful launch.

How these topics relate to Sys Admins:
Sys Admins do everything including DevOps. At the talk, I then learned that
DevOps don't know what they do themselves. Sys Admins deal with
filesystems, software installations, users, networking, dns/http,
monitoring, configuration, and far more. Since Kubernetes is a system for
automated deployment, scaling, and managing containerized applications, I
thought this overlapped nicely with some of the things Sys Admins need to
do on a daily basis.
In addition, deployment involves lots of configuration, networking, users,
etc. I was hoping that this talk would give useful information to manage
large scale installations for multiple users across multiple machines.

The speakers in each of the presentations were great. However, I felt that
the material could have been better. The first talk was a generic high
level view of what Kubernetes solves. DevOps are handle many complex
configurations and scripts in order to deploy software. They solve "But it
works on my machine ¯\_(ツ)_/¯". I thought this was similar to how system
administrators need to adapt to multiple environments and machines when
installing applications. Kubernetes could work
with many different types of systems and assists with minimizing
configuration for deployment. It also offers load balancing, server
discovery, scheduling, resource management, monitoring, and cloud
persistence. The best part is that these are consistent between
environments and machines. As mentioned in the other email, the speaker
discouraged setting up their own Kubernetes environment, but rather use 3rd
party provider such as the one his company provides.

The second talk was much more entertaining. The speaker talked about
deployments on April Fools, but the talk could actually be applied to all
deployments. Some points I got out of the presentation were:
- Swapping deployments to fix a bug is not a good idea. This can introduce
more bugs, downtime, and issues.
- Feature toggles should always be used to easily enable and disable a
particular feature. One of the problems with new features is self-inflicted
denial of service attacks. Users may get excited about new features, and a
company may not be prepared to handle a large amount of users. A feature
toggle can be temporarily used to disable the feature, scale, and then
re-enable the feature.
- I personally have not heard about dark launches before, but they're
great. The idea is to secretly deploy code that runs in the background of
an already deployed application. It can be used to collect extra
information or even deploy the feature, but hide it from most users. Google
did a dark launch on their homepage that ran a small javascript script to
test compatibility of IPv6 on user's computers.


- Brian Zawisza
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.stevens.edu/pipermail/cs615asa/attachments/20180321/aef5a903/attachment.html>