starred/netboot.xyz

Fork 0

mirror of https://github.com/netbootxyz/netboot.xyz.git synced 2026-04-25 23:25:54 +03:00

[GH-ISSUE #1203] Harvester install doesn't work; kind of an upstream problem #335

New issue

Closed

opened 2026-02-27 14:51:23 +03:00 by kerem · 3 comments

kerem commented

2026-02-27 14:51:23 +03:00

Owner

Originally created by @vincemulhollon on GitHub (Feb 16, 2023).
Original GitHub issue: https://github.com/netbootxyz/netboot.xyz/issues/1203

Describe the bug
The Harvester v1.1.1 install is incomplete and thus does not work.

To Reproduce
Steps to reproduce the behavior:

Install Harvester v1.1.1
Superficially installation appears successful. The system will reboot and report "Setting up Harvester" in the console. That will never complete and Harvester will not work.
If you log in to the console command line and poke around with journalctl and run various kubectl commands, the system seems permanently stuck in a loop where it run bootstrapmanifests/rancherd.yaml every five seconds which promptly errors out, repeat forever. If you look at the kubectl pods it seems the harvester-cluster-repo container is perma-stuck in ImagePullBackoff mode.

Expected behavior
A complete and operational Harvester installation.

Alternatively, you may want to temporarily flag the Harvester install menu option as not entirely working until the issue reports from the Harvester project linked below are fixed, after which the install menu option should start working. I think there is a non-default way to make it work although I am not sure what that is, so it would be bad to entirely delete the menu option.

Screenshots
Naw not really.

Additional context
As I (partially) understand the Harvester installation process, the cdrom/USB ISO image includes an additional partition full of containers used for "air gapped" installs or to prime the pump on internet connected installs to speed installation. As I understand the situation every container on that partition EXCEPT for harvester-cluster-repo is available online, so when netboot.xyz menu option for Harvester does NOT by default pass that pump-priming-partition to the Harvester install, every container except harvester-cluster-repo is slowly downloaded from the internet on initial boot. Unfortunately harvester-cluster-repo is rather critical to successful operation and as such the install will permanently fail at "Setting up Harvester" stage where a loop will run bootstrapmanifests/rancherd.yaml every five seconds forever. It seems harvester-cluster-repo is dynamically generated artifact of the ISO generation script or something like that and as such its not published online, intentionally. The Harvester project has a couple issues open on the overall situation and I interpret their long term solution is to publish a bootstrap-able harvester-cluster-repo container so installs would work even if the bootstrapping partition is never properly installed (it'll just download every container from the internet, and if in the future all of them are online, then that strategy will work)

A large collection of problems including this specific problem and some theoretical workarounds
https://github.com/harvester/harvester/issues/2651

A task from last August to permanently resolve the problem by publishing a copy of the container
https://github.com/harvester/harvester/issues/2670

Apparently there is a non-default process that I have not entirely figured out yet where a generic PXEBoot install (NOT using netboot.xyz) can send that bootstrapping partition full of containers to the new installs or the new installs can be pointed to some file store holding the harvester-cluster-repo container or ... something I don't entirely understand. Anyway my point is its believed in theory that its currently possible to PXEboot Harvester, but Harvester is working on a more straightforward way such that the default method would in the future actually work.

Originally created by @vincemulhollon on GitHub (Feb 16, 2023). Original GitHub issue: https://github.com/netbootxyz/netboot.xyz/issues/1203 **Describe the bug** The Harvester v1.1.1 install is incomplete and thus does not work. **To Reproduce** Steps to reproduce the behavior: 1. Install Harvester v1.1.1 2. Superficially installation appears successful. The system will reboot and report "Setting up Harvester" in the console. That will never complete and Harvester will not work. 3. If you log in to the console command line and poke around with journalctl and run various kubectl commands, the system seems permanently stuck in a loop where it run bootstrapmanifests/rancherd.yaml every five seconds which promptly errors out, repeat forever. If you look at the kubectl pods it seems the harvester-cluster-repo container is perma-stuck in ImagePullBackoff mode. **Expected behavior** A complete and operational Harvester installation. Alternatively, you may want to temporarily flag the Harvester install menu option as not entirely working until the issue reports from the Harvester project linked below are fixed, after which the install menu option should start working. I think there is a non-default way to make it work although I am not sure what that is, so it would be bad to entirely delete the menu option. **Screenshots** Naw not really. **Additional context** As I (partially) understand the Harvester installation process, the cdrom/USB ISO image includes an additional partition full of containers used for "air gapped" installs or to prime the pump on internet connected installs to speed installation. As I understand the situation every container on that partition EXCEPT for harvester-cluster-repo is available online, so when netboot.xyz menu option for Harvester does NOT by default pass that pump-priming-partition to the Harvester install, every container except harvester-cluster-repo is slowly downloaded from the internet on initial boot. Unfortunately harvester-cluster-repo is rather critical to successful operation and as such the install will permanently fail at "Setting up Harvester" stage where a loop will run bootstrapmanifests/rancherd.yaml every five seconds forever. It seems harvester-cluster-repo is dynamically generated artifact of the ISO generation script or something like that and as such its not published online, intentionally. The Harvester project has a couple issues open on the overall situation and I interpret their long term solution is to publish a bootstrap-able harvester-cluster-repo container so installs would work even if the bootstrapping partition is never properly installed (it'll just download every container from the internet, and if in the future all of them are online, then that strategy will work) A large collection of problems including this specific problem and some theoretical workarounds https://github.com/harvester/harvester/issues/2651 A task from last August to permanently resolve the problem by publishing a copy of the container https://github.com/harvester/harvester/issues/2670 Apparently there is a non-default process that I have not entirely figured out yet where a generic PXEBoot install (NOT using netboot.xyz) can send that bootstrapping partition full of containers to the new installs or the new installs can be pointed to some file store holding the harvester-cluster-repo container or ... something I don't entirely understand. Anyway my point is its believed in theory that its currently possible to PXEboot Harvester, but Harvester is working on a more straightforward way such that the default method would in the future actually work.

kerem

2026-02-27 14:51:23 +03:00

closed this issue
added the
bug
label

kerem commented

2026-02-27 14:51:24 +03:00

Author

Owner

@antonym commented on GitHub (Feb 18, 2023):

I know there were some issues on cluster startup I was seeing as well, but it seemed to also happen in the non netboot.xyz network install as well. I can probably tag it with experimental for now until it stabilizes. I'll try and find some time to take a look at it again. If you discover anything in the meantime, feel free to update it here or propose some changes. I know its changed over the releases so could be something we are missing still.

@antonym commented on GitHub (Feb 18, 2023): I know there were some issues on cluster startup I was seeing as well, but it seemed to also happen in the non netboot.xyz network install as well. I can probably tag it with experimental for now until it stabilizes. I'll try and find some time to take a look at it again. If you discover anything in the meantime, feel free to update it here or propose some changes. I know its changed over the releases so could be something we are missing still.

kerem commented

2026-02-27 14:51:24 +03:00

Author

Owner

@djieno commented on GitHub (Feb 9, 2024):

I can get a manual installation to work using pxe/netboot.xyz. That is already great !

I would like to use the config-create.yaml or config-join.yaml but it never gets resolved. I've set an nginx container with the config-create.yaml and config-join.yaml in the same subnet but no nginx log of netboot.xyz is using this during boot.

Set config-create or config-join.yaml URL: http://10.1.1.11:8888/harvester/config-create.yaml

Anyone got this to work ?

I need it since current pxe/netboot.xyz doesn't allow any pw to enter during the harvester installation. I need to set it using the create-x.yaml.

@djieno commented on GitHub (Feb 9, 2024): I can get a manual installation to work using pxe/netboot.xyz. That is already great ! I would like to use the config-create.yaml or config-join.yaml but it never gets resolved. I've set an nginx container with the config-create.yaml and config-join.yaml in the same subnet but no nginx log of netboot.xyz is using this during boot. Set config-create or config-join.yaml URL: http://10.1.1.11:8888/harvester/config-create.yaml Anyone got this to work ? I need it since current pxe/netboot.xyz doesn't allow any pw to enter during the harvester installation. I need to set it using the create-x.yaml.

kerem commented

2026-02-27 14:51:24 +03:00

Author

Owner

@goproslowyo commented on GitHub (Dec 9, 2024):

Would like to see https://github.com/netbootxyz/asset-mirror/releases/tag/v1.4.0-c82c6d22 included in a new netboot.xyz menu release :)

@goproslowyo commented on GitHub (Dec 9, 2024): Would like to see https://github.com/netbootxyz/asset-mirror/releases/tag/v1.4.0-c82c6d22 included in a new netboot.xyz menu release :)

kerem referenced this issue

2026-03-01 18:35:23 +03:00

[GH-ISSUE #335] Manjaro 18.0.4 - out of memory #1675

kerem referenced this issue

2026-03-01 18:35:26 +03:00

[GH-ISSUE #369] Add option for Manjaro Architect #1684

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

starred/netboot.xyz#335

No description provided.

Rows
Columns