Commit talk

Author: andrewdotn

presentation/docker-k8s-connection.pdf

@@ -100,17 +100,295 @@
 \tableofcontents
 \end{frame}
 
-\section{}
+\section{Docker}
+
+\begin{frame}{What is docker?}
+Problem: It’s hard to build and run software
+\pause
+\begin{itemize}
+\item getting it to build in the first place
+\pause
+\item having all required dependencies installed
+\pause
+\item at the right version
+\pause
+\item on different OS versions / distributions
+\end{itemize}
+\end{frame}
+
+\begin{frame}{Docker is:}
+Problem: It’s hard to build and run software
+\\[\baselineskip]
+Solution: Use the Linux system-call interface as the lowest-common
+denominator
+\pause
+\begin{itemize}
+\item No built-in applications, not even \texttt{ls}
+\item No built-in libraries
+\end{itemize}
+\pause
+\textit{All} dependencies must be packed into the container image
+\end{frame}
+
+\begin{frame}
+\includegraphics[height=0.95\paperheight,center]{what-you-ship1.pdf}
+\end{frame}
+
+\begin{frame}{Advantages}
+\begin{itemize}
+\item Speed: No more building code you didn’t write
+\pause
+\item Consistency: What you develop and test on your machine is
+\textit{exactly} what runs in production
+\pause
+\item Flexibility: Your app can run unmodified on any recent version of any
+Linux distribution like Ubuntu or CentOS
+\end{itemize}
+\end{frame}
+
+\begin{frame}[fragile]{Example}
+
+\begin{verbatim}
+docker run --rm -it centos
+# yum --version
+docker run --rm -it ubuntu
+# apt --version
+\end{verbatim}
+
+\pause
+
+\only<2>{
+\begin{itemize}
+\item \texttt{--rm}: remove after exit, for temporary containers
+\item \texttt{-i}: interactive, so open stdin
+\item \texttt{-t}: connect a terminal instead of a simple in-out stream
+\end{itemize}
+}
+
+\pause
+
+\begin{verbatim}
+docker run --rm --name my-tmp-db \
+    -e MYSQL_ROOT_PASSWORD=x mysql
+
+docker run --rm -it --link my-tmp-db \
+    mysql \
+    mysql -h my-tmp-db -px
+\end{verbatim}
+
+\end{frame}
+
+\begin{frame}{Big benefit for everyone}
+\sizefont{6}
+If you ever need to run a database or some other application for
+development purposes, it’s faster and easier to use docker
+\end{frame}
+
+\section{Kubernetes}
+
+\begin{frame}
+\tableofcontents[currentsection]
+\end{frame}
+
+\begin{frame}{What is kubernetes?}
+Problem: It’s hard to manage distributed systems
+\\[\baselineskip]
+\pause
+\only<2>{(A distributed system is a piece of software that requires more
+than one computer to run)}
+\pause
+\begin{itemize}
+\item Differences between running on a developer machine vs. running in a
+cloud or datacenter
+\item How do the pieces find each other?
+\item What happens when pieces fail?
+\item How do you upgrade?
+\end{itemize}
+\end{frame}
+
+\begin{frame}{Kubernetes is:}
+Problem: It’s hard to manage distributed systems
+\\[\baselineskip]
+Solution: Uniform APIs around implementations of good practices
+\end{frame}
+
+\begin{frame}{Docker and Kubernetes}
+\includegraphics[width=0.8\paperwidth,center]{docker-k8s-connection.pdf}
+\end{frame}
+
+\begin{frame}{Example of a good practice}
+Problem: A computer can physically fail, taking down the program running
+on it
+\\[\baselineskip]
+Solution: \\
+\only<2-5>{
+\begin{itemize}
+\pause
+\item Run copies on different computers
+\pause
+\item When a machine fails, create new copies of the programs on different
+computers
+\pause
+\item Redundant load balancers so clients don’t have to know about machine
+changes
+\pause
+\item Figure out some way to make the clients be ok with the load-balancers
+changing
+\end{itemize}
+}
+\pause
+That gets tricky to implement! You can use the standard kubernetes
+implementations of Service + Deployment instead.
+\end{frame}
+
+\begin{frame}{Deployment}
+Run $n$ copies of a container
+\pause
+\begin{itemize}
+\item If a container crashes, restart it
+\item If a computer crashes, restart the containers on a different one
+\item Give me a knob to easily adjust how many copies there are
+\item Give me an easy way to do rolling upgrades too
+\end{itemize}
+\end{frame}
+
+\begin{frame}[fragile]{Deployment}
+\sizefont{2}
+\begin{verbatim}
+apiVersion: extensions/v1beta1
+kind: Deployment
+metadata:
+  name: timestamp
+spec:
+  replicas: 1
+  template:
+    metadata:
+      labels:
+        app: timestamp
+    spec:
+      containers:
+      - name: hello
+        image: alpine/socat
+        command: ["socat", "-v", "TCP4-LISTEN:1234,fork,reuseaddr",
+          'SYSTEM:echo "Timestamp service: $(date)"']
+\end{verbatim}
+\end{frame}
+
+\begin{frame}{Service}
+Create a named thing that sends traffic to specific containers
+\pause
+\begin{itemize}
+\item Shows up in DNS inside cluster
+\item Fault-tolerant
+\item Updates dynamically
+\end{itemize}
+\end{frame}
+
+\begin{frame}[fragile]{Service}
+\sizefont{2}
+\begin{verbatim}
+apiVersion: v1
+kind: Service
+metadata:
+ name: timestamp
+ labels:
+   app: timestamp
+spec:
+  type: NodePort
+  ports:
+  - port: 1234
+    name: timestamp
+  selector:
+    app: timestamp
+\end{verbatim}
+Run \texttt{kubectl explain} for more details
+\end{frame}
+
+\begin{frame}[fragile]{Demo}
+\begin{verbatim}
+$ kubectl apply -f ../src/timestamp-svc.yaml
+
+$ kubectl get all
+
+$ kubectl run --rm --restart=Never -it \
+    --image centos test-shell
+[root@test-shell /]# curl timestamp:1234
+[delete pod]
+[root@test-shell /]# curl timestamp:1234
+
+\end{verbatim}
+\end{frame}
+
+\section{Driving Kubernetes with Python}
+
+\begin{frame}
+\tableofcontents[currentsection]
+\end{frame}
+
+\begin{frame}{APIs}
+\begin{itemize}
+\item Both docker and kubernetes are HTTP APIs under the hood
+\item \texttt{curl --unix-socket /var/run/docker.sock 'localhost/v1.37/containers/json'}
+\pause
+\item kubectl essentially converts YAML to JSON and POSTs it
+\pause
+\item You can create and test distributed computing entities on your laptop
+using the same API as what runs in production
+\end{itemize}
+\end{frame}
+
+\begin{frame}{Python code time}
+\end{frame}
+
+\section{Takeaways}
+
+\begin{frame}
+\tableofcontents[currentsection]
+\end{frame}
+
+\begin{frame}{Takeways}
+\begin{itemize}
+\item Whenever installing a database or application for development
+purposes, it’s probably faster and easier to use docker
+\pause
+\item If there are multiple moving pieces to your app, kubernetes can give
+uniformity across dev, test, and production
+\end{itemize}
+\pause
+Questions?
+\end{frame}
 
 \section{Exercises}
 
+\begin{frame}
+\tableofcontents[currentsection]
+\end{frame}
+
 \begin{frame}{Exercises}
 \begin{itemize}
-\item ...
+\item Install docker and run a container
+\item Run a container in kubernetes
+\item Run the example code that talks to kubernetes
 \end{itemize}
 Advanced:
 \begin{itemize}
-\item ...
+\item Build tooling for convenient local builds and runs of containers on
+kubernetes
+\end{itemize}
+\end{frame}
+
+\section{The Future of PyYYC}
+
+\begin{frame}
+\tableofcontents[currentsection]
+\end{frame}
+
+\begin{frame}{The Future of PyYYC}
+\begin{itemize}
+\item Summer hiatus
+\item Not 100\% sure on coming back in the fall
+\item The only significant way you can help: email [email protected]
+with talk abstracts
 \end{itemize}
 \end{frame}