Legacy packages have gone. How to update from 1.21?

General Discussions
1

Cluster information:

Kubernetes version: 1.21.10-00
Cloud being used: (put bare-metal if not on a public cloud) bare-metal
Installation method: kubeadm
Host OS: Ubuntu 20.04.6 LTS
CNI and version: containerRuntime=“containerd” version=“1.6.24” apiVersion=“v1alpha2”
CRI and version: I believe this was introduced after 1.21.

I didn’t see the deprecation blog post until today, and it seems the legacy packages have already gone. Is there any way to update, or should I rip our everything and start from scratch?

2

They shouldnt be gone yet, but will sometime this month.

FWIW - When I managed a bunch of clusters, I didn’t bother with the packages. They pretty much wrap a single binary, and a few config files. We would sync everything we needed to a fileshare and had a small bash script to do the install.

1 Like
3

honestly this really sucked this weekend, after about 4 years (yeah I know) my mc died in my pi cluster (4x pi 3b+) I know I should have moved etcd to an external drive but I didn’t :slight_smile: because its a learning cluster. Now that the legacy binaries have gone I am forced to move from pi 3+ to pi 5 (x4) really this has other implications in that pi5 and pi4 have higher power consumption and I can’t backfeed from usbhub anymore, this decision to drop direct support for armhf is understandable, but to loose the historical binaries absolutely sucks and forces me, and others I know, to spend hundreds on new kit OR move to vms (not everyone wants to run k3s guys).

Here endeth the lesson on relying on packages built and cached by others, I don’t do it at work, I was doing it at home and I should have just cached all the packages, I could rebuild the binaries, but yolo !

4

fwiw - all the binaries are still present. It’s just the packages that are gone. The packages mostly wrapped a single binary that made it easy to install using packaging tools, but you can also download them directly and run without too much effort (except for armhf^^;;). When I was a sysadmin with a bunch of clusters id download to a file share and had a bash script that would copy them out to all other nodes and cycle the upgrade.