Salman Raza:
Hi,
Some queries regarding the cluster upgrade question from exam point of view. Kindly help.
1.- As Skipping MINOR versions when upgrading is unsupported, so in exam if our current version( for example) is 1.26.4 and we need to upgrade to 1.27.3, then do we need to upgrade step by step, means first from 1.26.4 to 1.26.5 , 1.26.6 …1.27.3. OR Kubeadm upgrade apply wil automitically upgrade step by step from 1.26.4 to 1.27.3.
2.- During upgrade do we need to remove taint from nodes/control-plane so while draining the PODs scheduled to other nodes.
3.- In the previous documentation there was command "kubeadm upgrade node " for the Worker Nodes as well, but it is only mentioned for "Other Control Plane node in the latest docuemntation. Do we need to execute “kubeadm upgrade node” on worker nodes as well.
Thanks,
Salman
Salman Raza:
Kindly help on this.
Alistair Mackay:
1.
The major version is 1 (there is no version 2 in the pipeline currently)
The minor version is 25, 26 , 27 etc
The last number is “point release”
You can skip any number of point releases, therefore you can go direct from 1.26.4 to 1.27.3 in one step.
-
Depends if the question tells you that a pod or pods should end up on the control plane node following upgrade. For that to happen you would have to untaint. If there is no such instruction, simply drain the nodes. If the pods cannot schedule, they will remain in pending state until the node is uncordoned following upgrade and then will re-launch on the node.
-
kubeadm upgrade node is used to upgrade nodes of all types, except the first control plane node, which will receive new static pod manifests and images. Other control planes and all workers use upgrade node. There are currently no multi-master clusters in the exam environment.