cobbler/community

Roadmap: https://github.com/cobbler/cobbler/wiki/Roadmap Meeting Room: https://meet.opensuse.org/cobbler

zparihar

@zparihar

Hmmm... Damn.

I'm running a startup (in the early phase) and building an application based on Cobbler and I want/need Cobbler to manage my infrastructure. Currently it provisions only Red Hat based distros, but i'd like it to provision Debian (hence my questions earlier) and Ubuntu 22.04 as well. I think Debian will be straight forward to do within Cobbler.

I 'd like to engineer a solution somehow. I'm wondering about having a separate PXE TFTP server, and wonder if DHCP can point to that as the "next server" based on a MAC address.

I'm also wondering if I can raise some money from a Venture Capitalist and sponsor a volunteer in the community to develop this code and put it into Cobbler. I'm wondering how much I would need to sponsor someone?

ds-04

@ds-04

that is common, DHCP server and separate PXE server doing TFTP

ds-04

@ds-04

debian will work, i have it working via that ansible.. ive just not sorted out the post install phase fully (preseed post install).. which is easy with RHEL based and kickstart

1 reply

egotthold

@egotthold:matrix.org

@zparihar: Well the problem is not money. The problem is time. Sponsoring a single feature is nice but the person needs to get familiar with the code of Cobbler. Since Cobbler is not a small/easy codebase the biggest time of this job would be probably to do that. The feature itself is not that big since the issue linked above is already very well defined. The people working actively with the code should know already what to do and how. I don't have a perfect solution for the sitiuation at hand.
To get back to your problem: I would say that getting into the codebase that deep would take approx. two weeks. Implementing the feature after that time should be another week. However that assumes that the person is senior in Python and knows what it is doing. Feature of that size took in the beginning multiple months for me. I could probably pull a first buggy version in a weekend but my weekends are currently fully booked by my social obligations.

zparihar

@zparihar

@ds-04 I used to do Pre-Seeds with Debian probably about 10 years go with a good post install script. I'll look to get it working and will post my results.

@egotthold:matrix.org Thanks for the information. I think an about 6 months to a year when my company is in a good spot financially, I'd love to chat with you about this. By the way, are you 'schoolguy' on GitHub? If you are, I believe i've already had a conversation with you about a year ago.

egotthold

@egotthold:matrix.org

@zparihar: Yes I am. This is my Matrix Account. Sadly I am not aware on how to sync those two Accounts. Yes I vaguely remember the conversation we had.

jerelev

@jerelev

Hi, just installed cobbler 3.2.0 from sources and i get the following error in the cobbler_web:
In template /usr/share/cobbler/web/templates/generic_list.tmpl, error at line 222
Invalid block tag on line 222: 'ifequal', expected 'endblock'. Did you forget to register or load this tag?

Anyone knows why this happen ?
Django Version: 4.0.5
Exception Type: TemplateSyntaxError
Exception Value:
Invalid block tag on line 222: 'ifequal', expected 'endblock'. Did you forget to register or load this tag?
Exception Location: /usr/local/lib/python3.8/dist-packages/django/template/base.py, line 558, in invalid_block_tag
Python Executable: /usr/bin/python3
Python Version: 3.8.10
Python Path:
['/usr/share/cobbler/web/cobbler_web',
'/usr/share/cobbler/web',
'/usr/lib/python38.zip',
'/usr/lib/python3.8',
'/usr/lib/python3.8/lib-dynload',
'/usr/local/lib/python3.8/dist-packages',
'/usr/lib/python3/dist-packages']

egotthold

@egotthold:matrix.org

@jerelev: The old webinterface was not maintainable enough. If you open a bug maybe someone from the community that uses it may be able to help you. It could also be that Django 4.x is incompatible as I believe we only implemented it with Django 3.x. However I never touched that part. A new webinterface is under way but progress is very very slow due to the number of people working on Cobbler.

1 reply

Andrew Stephenson

@tendonut

hey folks, setting up a POC for Cobbler 3.4, TFTP seems to not be listening on 69/udp. settings.yaml is set to have cobbler manage it with in_tftpd, cobbler check comes back without tftp complaints, but ss/lsof is showing nothing listening on 69.

I used to run cobbler 2.8 when xinetd was involved, but that doesn't seem to be the case anymore.

Andrew Stephenson

@tendonut

A potential clue I'm finding.

cobbler-3.3.3-1.el9

[root@cobbler-poc01 ~]# cobbler validate-autoinstalls
task started: 2022-11-04_192120_validate_autoinstall_files
task started (id=Automated installation files validation, time=Fri Nov 4 19:21:20 2022)
running python triggers from /var/lib/cobbler/triggers/task/validate_autoinstall_files/pre/
running shell triggers from /var/lib/cobbler/triggers/task/validate_autoinstall_files/pre/
shell triggers finished successfully
validate_autoinstall_files
grab_tree found 3 children (including settings) of this object

osversion: rhel9
Exception occurred: <class 'TypeError'>
Exception value: unhashable type: 'Profile'
Exception Info:
!!! TASK FAILED !!!

there is no "validate_autoinstall_files" directory in /var/lib/cobbler/triggers/task
Looking back at cobbler/cobbler#2260, seems like this was a problem a while ago, but the bug report was closed.

zparihar

@zparihar

@tendonut Double check if the TFTP service is running.

Andrew Stephenson

@tendonut

Is it suppose to be a systemd unit? There is definitely no tftp service or process running.

Im not sure how the service is suppose to start, since cobbler is suppose to be "managing it" with that python module. And its not logging any errors regarding tftp.

javier-angulo

@javier-angulo

@tendonut , you need to install tftp server and enable service (or socket). On my centos8 box:

# dnf -y install tftp-server
# systemctl enable tftp.socket
# systemctl start tftp.socket

Andrew Stephenson

@tendonut

That did it. I was under the impression tftp didnt need to be managed by systemd because Cobbler was going to do it for me. I guess the documentation wasn't very clear on that.

Andrew Stephenson

@tendonut

Another weird issue. after installing "grub2-efi-x64-modules" and running cobbler mkloaders, its generating a grubx86.efi file, not an expected grubx64.efi file. The default value in settings.yaml is set to grubx86.efi, but the documentation is very specific about using the x64 version.

zparihar

@zparihar

@tendonut I remember experiencing something like that as well. If I recall, i got it all sorted out.

In the next 3 weeks, i'll be re-basing my product on Cobbler 3.2 (since its the most stable one). I'll publish my setup guide for Alma Linux 9 (because i'm rebasing my product on that too, lol).

Andrew Stephenson

@tendonut

that would be greatly appreciated. I'm standing up a POC for a replacement for our current bare metal provisioning solution internally. We as an org are FAR more familiar with Cobbler than any other solution since we've been using it for like 15 years lol

zparihar

@zparihar

@tendonut Amazing to hear. Whats the organization you're working for?

Andrew Stephenson

@tendonut

Red Hat lol. Customer Zero.

Andrew Stephenson

@tendonut

We rely heavily on satellite these days. We still have our old legacy cobbler systems running (2.6 i think), but they are used in our puppet-managed infrastructure. Everything now is Ansible/satellite, but we still need a more lean solution for purely bare metal provisioning we are discovering.

egotthold

@egotthold:matrix.org

@tendonut: So 3.4.0 is not yet released. That is the dev version essentially. RedHat was heavily involved in Cobbler before since it was the bare metal provisoning solution of choice for RH Satellite. There is an Ansible Community Module that is kinda working for 3.2.x and could deserve some love. If you need help for specific questions then feel free to open GH discussions so we can group your questions there.

@zparihar: If you rebase your product on 3.3.x at some point it would be good to hear what the issues are with it atm. Best collected in a GH Issue. I plan atm to reduce the open GH tickets with 3.4.x in a great deal and for this I would like a stable code base with few bugs...

Andrew Stephenson

@tendonut

It was actually 3.3.3-1 I was using, not 3.4.

I dont know why I said 3.4

im actually going OK now with a RHEL9 host and 3.3.3-1. I reprovisioned a new system that I havent been messing with and installation went pretty smooth.

egotthold

@egotthold:matrix.org

@tendonut: Perfect. Thanks for the reply. :)

zparihar

@zparihar

@egotthold:matrix.org I actually tested 3.3 a few months ago and I found a number of bugs, all of which I have reported. I haven't done a follow up.

zparihar

@zparihar

I know that there are some syntax changes between 3.2 and 3.3 so that sucks for my product development.

as i'll have to change the syntax at a later point when 3.3 has stabilized

teridon

@teridon

I'm trying to use cobbler-3.2.2 to PXE boot Ubuntu server focal (20.04) . Import seemed to work, and after cobbler sync there was a new PXE item for it in the menu. It looks like it starts booting, but then it starts outputting the error: /init: line 49: can't open /dev/sr0: No medium found

oh, wait I found this... cobbler/cobbler#2339

egotthold

@egotthold:matrix.org

@zparihar: Thanks for the update. I will try to keep that in mind.

@teridon: Yes Ubuntu is hard to use with Cobbler atm. I am starting to work on this, this weekend. :)

1 reply

egotthold

@egotthold:matrix.org

An update for the roadmap status from my side: https://cobbler.github.io/blog/2022/11/15/roadmap-update.html

1 reply

agraul

@agraul:opensuse.org

Hi, small question about the minimal Python version: Which one is correct 3.6 as stated in the wiki https://github.com/cobbler/cobbler/wiki/Code-Quality-Control or 3.10 as specified in https://github.com/cobbler/cobbler/blob/main/.pre-commit-config.yaml#L11?

egotthold

@egotthold:matrix.org

agraul: Python 3.6 is the correct one. The standard repositories of some distros don't include 3.6 anymore (like Tumbleweed) and this would break the pre-commit hook. Since we pinned the black version the checks should in theory stay consistent over the different Python versions afaik. Please correct me if I am wrong.

zparihar

@zparihar

@tendonut Satellite is a great product. Really have and really complicated. My product registers VM's in Katello (the lastest version) after its provisioned

@egotthold:matrix.org That's amazing (re: ubuntu) Let me know if you need some help testing. I just did a purchase order for 7 new servers and will be doing a lot of PXEBoot testing and Firewall testing with my Product and Cobbler integrations. I was about to do hack of PXE booting Ubuntu on a separate DHCP and TFTP server

egotthold

@egotthold:matrix.org

@zparihar: Perfect. Will ping you once I have a PR open. Since there were many discussions already that prepare the feature, I guess it will go rather sooner then later.

zparihar

@zparihar

excellent!

egotthold

@egotthold:matrix.org

@btravouillon: @zparihar I did not complete the cloud-init yet. I have merged a cleanup PR by myself which fixes roughly 1k pylint warnings instead. On Thursday I will be able to continue.

agraul

@agraul:opensuse.org

egotthold: how strongly do you care about keeping the file/mongodb serializers and all business logic in Python? Once item.to_dict is cached, a lot of time instantiating CobblerAPI is spent in System.find_match / System.find_match_single_key / System.__find_compare. I don't yet know the potential time saving, but I image this would run a lot faster in a relational database, e.g mariadb

zparihar

@zparihar

@egotthold:matrix.org No Problem

egotthold

@egotthold:matrix.org

agraul: So if we want to switch away from the JSON database we can do that. For 4.0 that was anyway the plan. I would prefer though if we can stay compatible also with sqlite as this is for smaller installations the "easiest" option imho. Then going for MariaDB and PostgreSQL is totally fine for me. A very early scratchpad for this in the past was the following issue: cobbler/cobbler#2725

agraul

@agraul:opensuse.org

I was actually thinking about SQLite, don't know why I wrote mariadb. Maybe we don't need to do the change yet anyway, I have a few other things I want to think about. Maybe we can improve some of the algorithms

egotthold

@egotthold:matrix.org

agraul: I saw your comments already on the Cache PR and Pablos experiments. However it never hurts to make something faster if possible... :)