[cs615asa] Meetup Summary

Divyendra Patil dpatil3 at stevens.edu
Fri Mar 23 13:59:49 EDT 2018


Meetup Date: March 20th, 2018
Name: DevOps for K8S + Operational Excellence in Apr Fools Pranks
More Info: https://www.meetup.com/nycdevops/events/246346445/

There were 3 meetups happening the same day in NY, but this one was more
aligned to System Administration and my primary reason to opt for the same.
Most of the people I met at the meetup were SysAdmin's, some from financial
firms in downtown NY while one had come all the way from kentucky.

There were two talks.

Talk 1:
By Paul Czakowski, An introductory presentation on Kubernetes, its
interests, and concerns of curious DevOps practitioners.

- Paul Czakowski says he is a DevOps guy in the day but he mocked that
himself doesn't know what his exact role is in the company he works for.
- Paul gave an overview of the components that make up Kubernetes and how
they work together to form a platform for scheduling and scaling containers.
- From Paul's talk, I had a brief idea of how kubernetes work and he
specifically differentiated how kubernetes are different from docker
containers, how they can be implemented and used in much more efficient
ways.
- From his talk, I can say that setting up kubernetes is no easy task since
there are 1000 things that can go wrong in doing the same. He himself
emphasized that we shouldn't do it and "IF" we are, should do it with the
guidance of a professional such as himself.
His statement "Companies like google are far better at handling and
managing Kubernetes than we will so don't do it. Out of 100, almost 97
people will fail at setting them up efficiently."
- He talked about making Kubernetes available to users and how they can be
composed together to deploy scalable and robust applications. He talked
about aspects of load balancing, monitoring, scheduling, resource
management, cloud, DNS configuration etc.
- He showed a lot of code of Json & YAML files for the configuration of
kubernetes and mentioned that it was fairly simple and possible to combine
multiple applications into one API by adding endpoints to the file.

The primary objects in Kubernetes include:
Pod, Service, Volume, Namespace. The various parts of the Kubernetes
Control Plane, such as the Kubernetes Master and kubelet processes, govern
how Kubernetes communicates with your cluster.
So when we use Kubernetes to create an object, we provide a new desired
state for the system. Kubernetes records that object creation and carries
out your instructions by starting the required applications and scheduling
them to cluster nodes–thus making the cluster’s actual state match the
desired state.

MORE:
I found a link which might briefly describe how Kubernetes work which might
be helpful:
https://blog.octo.com/en/how-does-it-work-kubernetes-episode-1-kubernetes-general-architecture/
Difference between Docker & K8: https://blog.containership.io/k8svsdocker

I have privately messaged Mr. Paul Czakowski for the slides but not sure
about his reply.


Talk 2:
By Tom Limoncelli, One of the highest stakes launches you can do is the
April Fools Prank (AFP).

Tom Limoncelli is a famous system administrator, author, and speaker at
many conferences around the globe on the topic of System Administration.
He was a systems administrator at Google & currently is an SRE (Site
Reliability Engineer) at Stack Overflow.

His talk was more light and entertaining compared to Paul.
In short, it was about how "NOT" to do things.

He shared many stories with an interesting one that happened at stack
overflow.
On April fool's day some time back, the people at stack overflow came up
with a simple game on their website to prank their users. The game got so
popular that it caused an uproar within 3 hours many users got addicted to
it which indirectly caused a self-inflicted DDoS attack on the servers, it
got so worse that even their load balancers were at full
capacity/overloaded.
It made the entire site unusable. They decided to bring down the feature
but changing it would take time and 3 minutes later the engineers involved
disabled the feature from the control panel and the entire system of stack
overflow was back to normal in some time.

Although his talk was more on the fun side, these are some points he
mentioned.
- It's better to hide a feature behind a flag than to roll out entire new
production to users.
- He made some points on Load Testing, (Everyone says they do it but nobody
actually does it).
His words, not mine. He emphasized on dark launches, he said that we can do
dark launches on some part of the users but nobody actually knows how it
will play out when it reaches to "all" the users.
- The concept is to have the feature running in the browser, but hidden
from users.
Google & Facebook do it for many products. Facebook did this before
deploying their chat (now called messenger). When they were going to roll
out this feature, they understood that it would suddenly be accessible to
millions of users and there was no way of testing it.
So what they did was, they called in a dark launch, 6 months before the
announcement, the code was in our browsers.
It just wasn't displaying the chat window to everyone & only certain
percentage of users were allowed access to it. They could send text
messages to other users and Facebook slowly kept cranking up the number of
users the feature was accessible to. This made them test the feature, find
out the bottlenecks and fix them before it was finally launched to everyone.

I think that these methods can and should be used by everyone in the
deployment phase of their product since it seems perfectly fine to test a
feature/product in phases rather than giving out to everyone just at one
instance.

He also showed us a small video as a joke:
https://www.youtube.com/watch?v=VgC4b9K-gYU

Since the first talk might not be understood just by audio, I have uploaded
the audio of the 2nd talk to my google drive if anyone is interested in
listening to it.
Here: https://drive.google.com/open?id=1mFC1M8b7wp5pLXW6NUKLL6IJqz-43HWF

You will find more info about Tom Limoncelli here:
https://en.wikipedia.org/wiki/Tom_Limoncelli




-- 
Thank You.
Divyendra Patil.
MS
Cyber Security
http://divyendra.com
‌
<https://mailtrack.io/> Sent with Mailtrack
<https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality&>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.stevens.edu/pipermail/cs615asa/attachments/20180323/b2d6fa14/attachment.html>


More information about the cs615asa mailing list