CRI-O

From ArchWiki
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.

CRI-O is an OCI-based implementation of the Kubernetes Container Runtime Interface.

As such it is one of the container runtimes that can be used with a node of a Kubernetes cluster.

Installation

Install the cri-o package.

The package will set the system up to load the overlay and br_netfilter modules and set the following sysctl options: 

 net.bridge.bridge-nf-call-iptables = 1
 net.bridge.bridge-nf-call-ip6tables = 1
 net.ipv4.ip_forward = 1

To use CRI-O without a reboot make sure to load the modules and configure the sysctl values accordingly.

Configuration

CRI-O is configured via /etc/crio/crio.conf or via drop-in configuration files in /etc/crio/crio.conf.d/.

Network

Plugin Installation

CRI-O can make use of container networking as provided by cni-plugins, or plugins installed with in-cluster deployments such as weave, flannel, calico, etc.

Warning: Arch installs the plugins provided by cni-plugins to both /usr/lib/cni and /opt/cni/bin but most other plugins (e.g. in-cluster deployments, kubelet managed plugins, etc) by default only install to the second directory.

CRI-O is only configured to look for plugins in the first directory and as a result, any plugins in the second directory are unavailable without some configuration changes.

This may present itself as a non-working network and in the CRI-O logs something like the following error will show: 

Error validating CNI config file /etc/cni/net.d/<plugin-config-file>.conf: [failed to find plugin "<plugin>" in path [/usr/lib/cni/]]

There are 2 solutions available to resolve this; have each of the other systems changed to use /usr/lib/cni instead or update CRI-O to use the latter directory instead of the first. The second solution can be achieved with a drop-in configuration file in /etc/crio/crio.conf.d/: 

00-plugin-dir.conf
[crio.network]
plugin_dirs = [
  "/opt/cni/bin/",
]

As this is an array, you can also set both or any other directories here as possible plugin locations.

Plugin Configuration

Copy one of the examples from /usr/share/doc/cri-o/examples/cni/ to /etc/cni/net.d and modify it as needed.

Warning: The cri-o package installs the 10-crio-bridge.conf and 99-loopback.conf examples to /etc/cni/net.d by default (as 100-crio-bridge.conf and 199-crio-loopback.conf respectively). This may conflict with Kubernetes cluster network fabrics (weave, flannel, calico, etc) and require manual deletion to resolve this (e.g. #2411 #2885).

Storage

By default CRI-O makes use of the overlay driver as its storage_driver for the container storage in /var/lib/containers/storage/. However, it can also be configured to use btrfs or ZFS natively by changing the driver in /etc/containers/storage: 

 sed -i 's/driver = ""/driver = "btrfs"/' /etc/containers/storage.conf

Running

Start and enable the crio.service systemd unit.

Testing

Use crio-status like this: 

 # crio-status info
 cgroup driver: systemd
 storage driver: vfs
 storage root: /var/lib/containers/storage
 default GID mappings (format <container>:<host>:<size>):
   0:0:4294967295
 default UID mappings (format <container>:<host>:<size>):
   0:0:4294967295

and: 

 # crio-status config
 ...

Now Install the crictl package, and see e.g. https://kubernetes.io/docs/tasks/debug-application-cluster/crictl/ or https://github.com/kubernetes-sigs/cri-tools/blob/master/docs/crictl.md, or simply: 

 source <(crictl completion bash)

 crictl pull index.docker.io/busybox
 crictl pull quay.io/prometheus/busybox
 crictl images

 curl -O https://raw.githubusercontent.com/kubernetes-sigs/cri-tools/master/docs/examples/podsandbox-config.yaml
 curl -O https://raw.githubusercontent.com/kubernetes-sigs/cri-tools/master/docs/examples/container-config.yaml
 crictl run container-config.yaml podsandbox-config.yaml

 crictl logs $(crictl ps --last 1 --output yaml | yq -r .containers[0].id)
 crictl exec -it $(crictl ps --last 1 --output yaml | yq -r .containers[0].id) /bin/sh

 crictl rm -af
 crictl rmp -af

Note how Docker Hub is not hard-coded, so specify container registry explicitly. (See also https://github.com/kubernetes-sigs/cri-tools/pull/718.)

See also