hpc-unibe-ch / pocket-ubelix2 Goto Github PK

Is your feature request related to a current shortcoming? Please describe.
Currently lrms01 and lrms02 are classified with the same role - slurmmaster -. To be able to
test the slurm-test cluster functionality, we need a slurmmaster_tester role

Describe the solution you'd like
Classify lrms02 with role slurmmaster_test.

Describe the bug
When running refresh_modules.sh the guard clause checking if we are in a VM
fails as the file it tests for is not available anymore.

To Reproduce

Steps to reproduce the behavior:

1. Create a fresh puppet master
2. Connect to the puppet master
3. Run /vagrant/refresh_modules.sh
4. See error

Expected behavior
Modules are refreshed.

Actual behavior
Script fails with error "Not in a virtualbox machine. Refusing to work."

Is your feature request related to a current shortcoming? Please describe.
Puppet modules delivered by VoxPupuli often require the gem toml-rb to be installed in
the puppetsever. The gem was manually installed on the production server.

See eMancu/toml-rb for the details about the gem.

Describe the solution you'd like
When running the script setup_puppetmaster.sh the gem toml-rb should be insatlled.

Is your feature request related to a current shortcoming? Please describe.
The puppet module puppet/grafana has a dependency on the toml rubygem - not toml-rb!

Describe the solution you'd like
Add gem toml to the puppetserver. Having to deal with toml and toml-rb now, also outline which puppet module requires which gem.

Describe the bug
When having a symlink from puppet env to a local module in development, running refresh_modules.sh
messes up my local repo as g10k does not kill the symlink and checkout a fresh working tree but instead
copies diretyl into the symlinked directory => messed up local repo.

To Reproduce

Steps to reproduce the behavior:

1. Go to local puppetmaster in pocket UBELIX
2. Add a symlink to a module currently in development:
   • rm -rf /etc/puppetlabs/code/environment/development/modules/slurm
   • ln -s /vagrant/controlrepo/modules/puppet-module-slurm /etc/puppetlabs/code/environment/development/modules/slurm
3. Make a change in the module
4. Run refresh_modules.sh
5. Verify that it messed with the module slurm.

Expected behavior
As a puppet engineer on the puppet master of the pocket UBELIX I want to have the
described symlinks to my local modules in devleopment without having the script
refresh_modules.sh messing up my local git repository.

Actual behavior
As g10k follows symlinks and has the impression that the module is already there, it just
copies the downloaded HEAD revision of the module and thus overwrites my local change in
the repositories.

Is your feature request related to a current shortcoming? Please describe.
The current Vagrantfile has options that are never used or repeated. In addition it uses option keys that are from APIv1 that are deprecated.

Describe the solution you'd like
Cleanup the Vagrant file and

• remove unsused or irrelevant optoins
• replace option key host_name with hostnme

Is your feature request related to a current shortcoming? Please describe.
As the user and group of slurm and munge are no longer managed by Puppet but stored in central
LDAP, we need to handle them in Pocket UBELIX anyway.

Describe the solution you'd like
Create these local users and groups at VM boot time by calling a script that creates them if
not already available.

Is your feature request related to a current shortcoming? Please describe.
The default setting for checking for a newer version of the base box a machie was built
on is true. That means on every machine startup this is checked and a lot of output is generated
if there is a newer base box. But there is no sense to recreate the machine only because there
is a newer base box, at least I don't do that.

Describe the solution you'd like
Turn off checking for newer base box.

Describe the goal of the task
Upon tuning the puppet servers and if it has been shown to work, remove the runinterval from the puppet.conf
of all clients in order to go back to a default runinterval of 1800 seconds.

Action items

Steps to carry out to fullfill the task:

• Lookup if a puppet agent restart is required after changing puppet.conf
• Change runinterval: pdsh -g puppetclients puppet config delete --section agent
• Remove the runinterval setting from setup_puppetmaster.sh and setup_puppet-agent.sh
• Update /usr/local/sbin/setup_puppet-agent.sh from github.com by downloading it again

Definition of done
All puppet agents run with an interval of 30m. Newly kickstarted servers are using the default of 30m.
All rolled out /usr/local/sbin/setup_puppet-agent.sh scripts are not setting runinterval anymore.

Describe the bug

After installing a new virtual box, the setup_puppetmaster.sh script breaks off when it is executed.

'set -e' Exit immediately if a command exits with a non-zero status.

The last change

"Install reuqired gem diff-lcs for show_diff"

ends with a non-zero status.

We decided not to change the query.

set -e is removed

Is your feature request related to a current shortcoming? Please describe.
Currently the show_diff functionality is not working apparently. Recently learnt that the
diff-lcs ruby gem is necessary for this feature:
https://puppet.com/docs/puppet/7.4/configuration.html#show_diff

Describe the solution you'd like
Deploy the diff-lcd withing the puppetserver when installing it.

Is your feature request related to a current shortcoming? Please describe.
See issues hpc-unibe-ch/ubelix-controlrepo#144 and hpc-unibe-ch/ubelix-controlrepo#145

Describe the solution you'd like
Decide which changes are necessary in the puppet installer to achieve the goal outlined in the above two issues.

Describe the bug
The previous PR was incomplete. The node definitino for lrms02 is invalid.

To Reproduce
Steps to reproduce the behavior:

1. Clone Puppet-UBELIX 2
2. Provision any host mention in Vagrantfile

Expected behavior
Any machine provisions as expected.

Actual behavior
It's not possible to provision any machine due to invalid Vagrantfile.

Vagrant failed to initialize at a very early stage:



There was an error loading a Vagrantfile. The file being loaded
and the error message are shown below. This is usually caused by
an invalid or undefined variable.



Path: /Users/Matthias/priv_puppet/pocket-ubelix2/Vagrantfile
Line number: 0
Message: undefined local variable or method `index'