This is a continuation of my previous blog on macvlan and ipvlan Linux network drivers. Docker has added support for macvlan and ipvlan drivers and its currently in experimental mode as of Docker release 1.11.
In this example, we will use Docker macvlan and ipvlan network plugins for Container communication across hosts. To illustrate macvlan and ipvlan concepts and usage, I have created the following example.
Following are details of the setup:
Macvlan and ipvlan are Linux network drivers that exposes underlay or host interfaces directly to VMs or Containers running in the host. In this blog, I will cover basics of macvlan and ipvlan, compare macvlan and ipvlan to Linux bridge and sub-interfaces and also show how to create these interfaces in Linux system. In the next set of blogs, I will cover how macvlan and ipvlan interfaces are used in Docker and CoreOS.
When running a baremetal server, host networking can be straightforward with few ethernet interfaces and a default gateway providing external connectivity. When we run multiple VMs in a host, it is needed to provide connectivity between VMs within the host and across hosts. On an average, the number of VMs in a single host does not exceed 15-20. When running Containers in a host, the number of Containers in a single host can easily exceed 100. It is needed to have sophisticated mechanism to interconnect Containers. Broadly, there are two ways for Containers or VMs to communicate to each other. In Underlay network approach, VMs or Containers are directly exposed to host network. Bridge, macvlan and ipvlan network drivers are examples of this approach. In Overlay network approach, there is an additional level of encapsulation like VXLAN, NVGRE between the Container/VM network and the underlay network.
Linux Bridge acts like a regular hardware switch with learning and also supports protocols like STP for loop prevention. In linux bridge implementation, VMs or Containers will connect to bridge and bridge will connect to outside world. For external connectivity, we would need to use NAT. The following picture shows 2 Containers connected to a Linux bridge with ethx interface providing external connectivity.
Docker Experimental channel is used to release experimental Docker features so that Docker users can try the new features and provide feedback. It is nice to use the experimental Docker in a test environment rather than upgrading Docker in the main development machine. The preferred approach is to use docker-machine and create a VM with experimental Docker. In this blog, I will describe the approach that I use to create docker-machine with experimental Docker VM. For basics of Docker machine, please refer to my blog on Docker machine.
Following are the steps needed to build the experimental boot2docker ISO and copy it to the docker-machine default location:
git clone https://github.com/boot2docker/boot2docker.git cd boot2docker docker build -t my-boot2docker-img -f Dockerfile.experimental . docker run --rm my-boot2docker-img > boot2docker.iso mv boot2docker.iso ~/.docker/machine/cache/boot2docker.iso
We need to specify Docker experimental release location in Dockerfile.experimental. In this case, it is https://experimental.docker.com/builds/Linux/x86_64/docker-latest.
Following command will start a Docker machine in Virtualbox with experimental Docker:
docker-machine create -d virtualbox exp
Following is the experimental Docker version running in my host:
$ docker --version Docker version 1.11.0-dev, build 6c2f438, experimental
Installing custom software in boot2Docker image:
I had a recent usecase where I needed boot2docker to have ipvsadm installed. Package manager is not available in boot2docker. Other than installing ipvsadm, I had to copy few libraries. Following is my boot2docker.experimental file that I used for this usecase:
FROM boot2docker/boot2docker MAINTAINER Sreenivas Makam "" #DESCRIPTION use the latest experimental build of Docker RUN apt-get update && apt-get install -y ipvsadm RUN cp /sbin/ipvsadm $ROOTFS/sbin/ RUN cp /lib/x86_64-linux-gnu/libnl-genl-3.so.200 /rootfs/lib/libnl-genl-3.so.200 RUN cp /lib/x86_64-linux-gnu/libnl-3.so.200 /rootfs/lib/libnl-3.so.200 RUN cp /lib/x86_64-linux-gnu/libpopt.so.0 /rootfs/lib/libpopt.so.0 #get the latest experimental docker RUN cd $ROOTFS/usr/local/bin && curl -fL -O https://experimental.docker.com/builds/Linux/x86_64/docker-1.12.0-rc4.tgz && tar -xvzf docker-1.12.0-rc4.tgz --strip-components=1 && chmod +x $ROOTFS/usr/local/bin/docker* && rm docker-1.12.0-rc4.tgz RUN echo "" >> $ROOTFS/etc/motd RUN echo " WARNING: this is an experimental.docker.com build, not a release." >> $ROOTFS/etc/motd RUN echo "" >> $ROOTFS/etc/motd RUN /make_iso.sh CMD ["cat", "boot2docker.iso"]
Issue faced:
I was not able to use custom Docker image with docker-machine version 0.8.0-rc1. I could not find an option to prevent docker-machine from downloading latest Docker image. I have opened an issue here. The only workaround I found was to copy boot2docker image to ~/.docker/machine/cache/ , remove internet connection and then create docker-machine host.
Contiv is an Open source project driven primarily by Cisco for policy based networking, storage and cluster management for containerized applications. In this blog, I will cover some of the hands-on stuff that I tried with Contiv Networking. I used the sample examples provided in Contiv documentation as starting point. For Contiv networking basics, you can refer to my previous blog here.
I followed the “Contiv getting started” guide to setup a two node Contiv cluster with Vagrant. I started the cluster in Packet baremetal cloud. Contiv netmaster runs in one of the nodes, Contiv netplugin is installed in both the nodes.
git clone https://github.com/contiv/netplugin cd netplugin; make demo
Following command shows the 2 node Vagrant cluster:
root@contiv:~/netplugin# vagrant status Current machine states: netplugin-node1 running (virtualbox) netplugin-node2 running (virtualbox)
Following are the business details of the sample application that I have used in this blog:
Contiv is an Open source project driven primarily by Cisco for policy based networking, storage and cluster management for containerized applications. In this blog, I will focus on how Contiv does policy based Container networking. In the next blog, I will cover some hands-on stuff that I tried with Contiv.
Policies have become critical to control the business logic in a Cloud environment. There are 2 ways to describe policy. In imperative model, policy is defined in terms of how the end goal is achieved. For example, we specify the filters and actions with Openflow protocol that achieves end goal of packet handling and this is an example of imperative model. In declarative model, policy is defined in terms of the end goal and it gives flexibility to the end-system to implement the policy in different ways. Congress and Opflex are examples of declarative policy model. With declarative model, it is possible to specify the policy in terms of business logic without specifying implementation detail. For example, the business logic can say that web container should not talk to database container. The implementation of this business logic can be achieved by having an iptables rule or by having a hardware tcam rule to block specific ports. In a cloud computing world, policies can be defined for compute, storage and networking. Both Containers and VM needs policies to implement business logic. Following are examples of some policies that can be applied to applications deployed in Cloud using either VMs or Containers:
Contiv Networking project provides policy based networking for Docker Containers. Following are some details on Contiv Networking:
Continue reading Contiv – Policy based networking for Containers
I recently saw the Openstack self-healing demo from CoreOS team using Tectonic(Stackanetes project) and I kind of felt that the boundary between Containers and VMs are blurring. In this blog, I discuss the usecase of deploying Openstack using Containers.
We typically think of Openstack as a VM Orchestration tool. Openstack is composed of numerous services and deploying Openstack as one monolithic blob is pretty complex and difficult to maintain. The demo described showed how Containers simplify Openstack deployment. This is a great example of using Microservices architecture to simplify infrastructure deployment.
Following diagram shows the Openstack deployment model using Containers. The diagram below shows how Openstack service containers deploys user VM. The user VMs deployed using Openstack can run Containers as well..
Following are some notes on the architecture:
This work also shows how Containers and VMs can work closely with each other for lot of use-cases. There are other Openstack projects like Magnum and Kuryr where there is an intersection between Containers and VMs. Magnum project deals with Container orchestration using Openstack and Kuryr project deals with doing Container networking using Openstack Neutron.
References:
I did a presentation on CoreOS and Service Discovery in Opensource Meetup group last week. Following are related slides and demo recording.
Scripts used are available here.
Following are the commands to start Consul Container, Registrator Container and 3 Container services.
docker run -d -p 8500:8500 -p 192.168.0.1:53:8600/udp -p 8400:8400 gliderlabs/consul-server -node myconsul -bootstrap docker run -d -v /var/run/docker.sock:/tmp/docker.sock --net=host gliderlabs/registrator -internal consul://localhost:8500 docker run -d -p :80 -e "SERVICE_80_NAME=http" -e "SERVICE_80_ID=http1" -e "SERVICE_80_CHECK_HTTP=true" -e "SERVICE_80_CHECK_HTTP=/" --name=nginx1 nginx docker run -d -p :80 -e "SERVICE_80_NAME=http" -e "SERVICE_80_ID=http2" -e "SERVICE_80_CHECK_HTTP=true" -e "SERVICE_80_CHECK_HTTP=/" --name=nginx2 nginx docker run -ti smakam/myubuntu:v3 bash
In a Microservices architecture, Services are dynamic, distributed and present in large numbers. It is needed to have a good Service discovery solution to address this dynamic problem. In this blog, I will cover basics of Service discovery and using Consul to do Service discovery.
Service discovery should provide the following:
Following are the critical components of Service discovery:
Mantl is an Open source project from Cisco and it provides an integrated solution to deploy distributed Microservices. Any company deploying Microservices has to integrate different components before the solution becomes production ready. Mantl makes it easier by integrating the different components and providing the glue software that integrates the components. In this blog, I will cover the following:
Following are typical components in Container based Microservices infrastructure:
This is the fourth part of my Docker security series. In this blog, I will cover ways to secure Container images. Following are the other parts (1, 2, 3).
The most critical part with Docker is the Container image as ultimately the platform is being used to develop and deploy Container images. Docker Container images can be stored either in public or private registry. It is needed to sign Container images so that the client knows that image is coming from a trusted source and that it is not tampered with. Content publisher takes care of signing Container image and pushing it into the registry.
Following are some details on Docker content trust:
Sreenivas Makam's Blog 