Setting up a new node in the cluster: Difference between revisions

From Collective Computational Unit
Jump to navigation Jump to search
← Older edit
undoBastian.goldluecke (talk | contribs)
ccu, Administrators
684 edits
mNo edit summary
Tag: visualeditor-switched
 
(5 intermediate revisions by the same user not shown)
Line 1: Line 1:
== Docker ==
== Preparing the node ==


* follow the steps to install docker, nvidia-docker2 and make nvidia-docker the default runtime on the node.
* Create new user "kubernetes" with groups docker and sudo.
* Follow the steps to install docker, nvidia-docker2 and make nvidia-docker the default runtime on the node.
* Follow instructions [[Initializing_the_Kubernetes_cluster|on this page]] to install Kubernetes and pre-requisites.


== Spin up the slave node ==


== Kubernetes ==
On the master node:


Follow instructions [[Initializing_the_Kubernetes_cluster|on this page]] to install Kubernetes.
<syntaxhighlight lang="bash">
> cd init
> ./show_join_command.sh
</syntaxhighlight>


== Post-install configuration ==
Copy the output into a new file "join_cluster.sh" in the init directory of the slave.


* Add the node to the different namespaces it should belong to.
On the slave:
** TODO
 
<syntaxhighlight lang="bash">
> cd init
> chmod +x join_cluster.sh
> sudo ./join_cluster.sh
</syntaxhighlight>
 
 
There are problems with the flannel network currently, which requires manual patching of the node after installation, and restarting the flannel pod:
 
<syntaxhighlight lang="bash">
> kubectl patch node <NODE_NAME> -p '{"spec":{"podCIDR":"10.244.0.0/16"}}'
</syntaxhighlight>
 
 
 
[[Category:Internal]]

Latest revision as of 17:00, 17 November 2020

Preparing the node

[edit]
  • Create new user "kubernetes" with groups docker and sudo.
  • Follow the steps to install docker, nvidia-docker2 and make nvidia-docker the default runtime on the node.
  • Follow instructions on this page to install Kubernetes and pre-requisites.

Spin up the slave node

[edit]

On the master node: 

> cd init
> ./show_join_command.sh

Copy the output into a new file "join_cluster.sh" in the init directory of the slave.

On the slave: 

> cd init
> chmod +x join_cluster.sh
> sudo ./join_cluster.sh


There are problems with the flannel network currently, which requires manual patching of the node after installation, and restarting the flannel pod: 

> kubectl patch node <NODE_NAME> -p '{"spec":{"podCIDR":"10.244.0.0/16"}}'
Retrieved from "https://ccu-wiki.inf.uni-konstanz.de/index.php?title=Setting_up_a_new_node_in_the_cluster&oldid=712"