k3s: Latest `install.sh` exits without error message on airgapped nodes

Environmental Info: K3s Version: master-branch

Node(s) CPU architecture, OS, and Version: Ubuntu 20.04 LTS

Cluster Configuration: N/A

Describe the bug: When running the latest install.sh with below command, it exits right after showing [INFO] Skipping k3s download and verify without an error message (exit code is not 0 though):

INSTALL_K3S_SKIP_DOWNLOAD=true INSTALL_K3S_EXEC="server --cluster-init --token <token> --tls-san <ipaddress> --disable local-storage --config /etc/rancher/k3s/config.yaml" ./install.sh

Steps To Reproduce: See above

Installed K3s: N/A

Expected behavior: Installation starts.

Actual behavior: See above

Additional context / logs: The install.sh from tag v1.26.3+k3s1 still works correctly.

/kind bug

About this issue

Original URL
State: closed
Created a year ago
Reactions: 1
Comments: 19 (7 by maintainers)

Most upvoted comments

I am seeing an error that may have the same root cause, related to using the -P option with grep. I started seeing this on Friday, April 7. The timing of when the error started seems to coincide with the code change referenced earlier in the thread.

root@qemuarm64:~# INSTALL_K3S_EXEC="--snapshotter=native" INSTALL_K3S_SKIP_DOWNLOAD=true ./k3s_install.sh
[INFO]  Skipping k3s download and verify
grep: invalid option -- 'P'
BusyBox v1.31.1 (2022-12-13 23:09:41 UTC) multi-call binary.

Usage: grep [-HhnlLoqvsriwFE] [-m N] [-A/B/C N] PATTERN/-e PATTERN.../-f FILE [FILE]...

jgbright on Apr 10, 2023

Not sure if this is realated but I am seeing today that curl -sfL https://get.k3s.io | bash - started to fail intermittently but its not exiting 1

When I SSH’d on to the EC2 instance I had to run the install script by hand a few times before it worked

I’m trying to work around it with some retries in my script, but if this is round robin’ing to some old script thats in cache or what not then retry’ing will start failing

echo "INSTALLING_K3S" > /home/admin/status

install_k3s() {
  echo "installing k3s"
  curl -sfL https://get.k3s.io | bash -
  sudo chown -R admin:admin /etc/rancher/k3s
  echo "finished installing k3s"
}

# Install K3s with 20 retries
for i in {1..20}; do install_k3s && break || sleep 1; done

echo "INSTALLING_ADMISSION_CONTROLLER" > /home/admin/status

I might look into https://github.com/k3s-io/k3s/issues/7247#issuecomment-1500711907 and see if I can pin to a known working script.

fieldju on Apr 7, 2023

We fell back to the previous version of the script for now, as we need to spin up a cluster for a competition this weekend (new cluster each week). Will look out for the fix on this ticket!

Link to previous version of script for anyone looking to do the same: https://raw.githubusercontent.com/k3s-io/k3s/7e59376bb91d451d3eaf16b9a3f80ae4d711b2bc/install.sh

Rudedog9d on Apr 7, 2023

@galal-hussein it looks like the change at https://github.com/k3s-io/k3s/commit/027cc187ce9f21157b8d37d62e67ee1c42968b4b#diff-043df5bdbf6639d7a77e1d44c5226fd7371e5259a1e4df3a0dd5d64c30dca44fR500-R504 makes it call out to the GitHub API to retrieve the latest release of the k3s-selinux package, even if it’s not going to be shown.

While I do like the idea of having the script call out to get the latest version to suggest, we probably need to make that an optional check with a short timeout, and fall back to a known-good default value. It also shouldn’t happen when we’re not downloading anything.

brandond on Apr 7, 2023