background-image: url(faces_shipping_containers.jpg)
.sigblock[ Josh Berkus
Red Hat
PDX DevOps 10/17 ]
"Your purpose should be always to know the whole that was intended to be known."
"A sensible man should not demand of me, or hope that when we mention a subject, we shall make a complete exposition of it."
.right-column[
]
.right-column[
* Linux Namespaces * Cgroups * Networking tricks ]
background-image: url(container_reflection.jpg)
- a tarball
containing a user FS - some metadata
image == packaging
- union file system:
overlayFS, AuFS, devicemapper - layers can be added
- layers can be shared
- Copy a base image ("Fedora 26")
- Add new layer(s)
- Add metadata
- Bundle up
- Docker Build (Dockerfile)
- rkt
- Buildah
- ad-hoc
- hub.docker.io
- Google container registry
- CoreOS quay.io
- Specialty: CentOS, Fedora, etc.
- local registry
- find on registies
- download to a local cache
- copy to memory at runtime
docker pull, skopeo
start a single binary from the image running on the host OS in a namespaced process, file, and resource space
- Docker Daemon
containerd - Rkt + Systemd
- OCI + Systemd
- CRI-O
background-image: url(many_containers.jpg)
- smaller
container support, nothing else - immutable
updates like containers
run many containers
over many machines
- Scheduler: dispatches containers to machines
- Agents: manages containers on each machine
- API Server: communicates between machines
- DCS: shared data store for cluster
- Plugins: storage, networking, etc.
out of the box, containers have only the host network, so we need to fake a network using Linux kernel tools and networking components so that they can pretend to be servers
finding your containers
- name
- service type
- port
some DNS, some not
- some containers need to store data
- data can be ephemeral
(local Volumes) - data can also be long-term
(Persistent Volumes) - sometimes data needs to move with the container
- network storage (NAS)
- local storage w/copying
- cloud storage (Gluster, Ceph, Rook, EBS)
no "container storage" ... yet
background-image: url(container_heap.jpg)
- CaaS: run containers on their infra (GKE, Amazon CS)
- PaaS: run applications as containers on their infra (OpenShift, DEIS)
- FaaS: run code functions as containers on their infra (Serverless)
- monitoring
- tracing
- CI/CD tools
- API tools
.left-column[
web:
events: ]
.right-column[ jberkus.github.io www.databasesoup.com www.projectatomic.io
Cloud Native Portland
meetup.com/Cloud-Native-PDX/
LISA - KubeCon
SCALE Container Day
]


















