This change is to update the documentation for reserved-memory flag.
Now, for specifying memory reservations across multiple NUMA nodes,
semicolon needs to be used as separator.
Signed-off-by: Ravindra Thakur <ravindra.nath.thakur@est.tech>
The command "kubeadm kubeconfig user" is missing
some examples and more details on why it is needed.
- Add a new section "Generating kubeconfig files
for additional users" under the kubeadm-certs page.
- Link to this section from the kubeadm-kubeconfig reference page.
- Link to this section from the create-cluster-kubeadm page.
Sysbox is an open-source container runtime (similar to "runc") that supports
running VM-workloads such as Docker and Kubernetes inside unprivileged
containers or pods.
Sysbox containers always use the Linux user-namespace for isolation, plus
specially crafted proc and sys filesystems, some syscall interception,
filesystem ID-mapping, and more.
It's possible to run Kubernetes, K3s, K0s, inside containers or pods deployed
with Sysbox. This commit aims to make the Kubernetes community aware of this
option.
Signed-off-by: Cesar Talledo <ctalledo@nestybox.com>
Touch the following files:
- Implementation details: remove docker specifics, which is changing
in 1.24
- Create cluster: small language cleanup, remove note about 1.24
- Install kubeadm: Include two up-to-date tables for Linux / Windows
with known endpoints. Include cri-dockerd.
- Kubelet integration: (side cleanup) use "container runtime" instead of
"CRI runtime" (which is incorrect). Mention that only updating
"--container-runtime-endpoint=.." is required if the user wishes
to override the CR on a certain host. Dockershim->CR-foo migration
guides would make the "--container-runtime=remote" flag explicit
and we want to remove it at some point.
- Troubleshooting kubeadm: Remove some instances of Docker troubleshooting
that imply docker as default CR, or talk about old Docker versions.
Be more generic about container runtimes.
- Adding Windows nodes: move the containerd tab before the Docker
tab, as containerd is now the default. Remove note about being explicit
about --cri-socket. Add note that crictl is required for both
Docker and containerd. Add note that cri-dockerd is required if
the user wants to use Docker EE on Windows.
- Fixed brain flatuence around which runtime I was testing
- Outputs changed
- Noted added about crictl ps's default behavior of outputting sha256sum in image name
- Image name in outputs have been shortened per Slack convo https://kubernetes.slack.com/archives/C02PZMA3R6E/p1643026840047600