Fix upgrade size calculation inside k8s

[Itxaka]

As we use the suc-upgrade script to upgrade inside k8s containers, the host rootfs is mounted under $HOST_DIR. The upgrade image fs is actually the rootfs of the container, so its mounted on /

The upgrade calculation was going over this dir when calculating the size for the upgrade, as the host dir is mounted under / This means that we were calling the upgrade with --source dir:/ and that tried to go into the /$HOST_DIR and count that size.

This created 2 issues. The first one is that the upgraed size would be doubled as it counted the upgrade image + the host contents. The second would be that counting an installed system is a bit problematic with dirs like /tmp or /proc in which there can be temp files, broken symlinks and symlinks to self which wont work inside the container (self links point to the running system, so a self link in the mounted host would point to a different process than in the running container)

To fix this, we try to get the $HOST_DIR variable when calculating the sizes. The suc script sets this ALWAYS so if we are running under k8s with an upgrade, we will get it and we can skip counting that dir.

KUBERNETES_SERVICE_HOST is alway set for pods running under k8s so we
can infer that we are under k8s if that its set.

If that its true and the HOST_DIR var is not set, we default to the safe
value that system-upgrade-controller sets the host root into

[Itxaka]

This requires backport to 2.13.x branch

[codecov]

Codecov Report

Attention: Patch coverage is 83.33333% with 2 lines in your changes missing coverage. Please review.

Project coverage is 49.90%. Comparing base (83afd40) to head (6f61df4).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
pkg/config/spec.go	83.33%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #537      +/-   ##
==========================================
+ Coverage   49.84%   49.90%   +0.05%     
==========================================
  Files          48       48              
  Lines        4580     4589       +9     
==========================================
+ Hits         2283     2290       +7     
- Misses       2028     2029       +1     
- Partials      269      270       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jimmykarily]

Good catch!

[Itxaka]

Tested with k8s and it works as expected!

k3s upgrade test from k8s installs to disk with custom config [provider, provider-upgrade-latest-k8s-with-kubernetes]
/home/itxaka/projects/kairos/tests/provider_upgrade_latest_k8s_test.go:59
State dir: /tmp/1292742850
Using ssh port: 36857
  STEP: checking if it has default service active @ 09/13/24 12:40:46.999
  STEP: copy the config @ 09/13/24 12:40:47.194
  STEP: find the correct device (qemu vs vbox) @ 09/13/24 12:40:47.282
  STEP: installing @ 09/13/24 12:40:47.359
  STEP: rebooting after install @ 09/13/24 12:41:04.496
  STEP: checking default services are on after first boot @ 09/13/24 12:42:04.706
  STEP: Checking agent provider correct start @ 09/13/24 12:42:05.177
  STEP: checking kubeconfig @ 09/13/24 12:42:05.242
  STEP: checking current version @ 09/13/24 12:42:05.3
  STEP: wait system-upgrade-controller @ 09/13/24 12:42:05.361
  STEP: wait for all containers to be in running state @ 09/13/24 12:42:05.361
  STEP: triggering an upgrade @ 09/13/24 12:42:15.746
  STEP: checking upgraded version @ 09/13/24 12:42:16.192
• [1021.403 seconds]
------------------------------
SS

Ran 1 of 21 Specs in 1021.404 seconds
SUCCESS! -- 1 Passed | 0 Failed | 0 Pending | 20 Skipped
PASS