Skip to content

Magic Castle 7.3

Compare
Choose a tag to compare
@cmd-ntrf cmd-ntrf released this 04 Jun 19:03
f69db7f

This release introduces three main features:

  • Add support for Slurm 20
  • Add support for CentOS 8. Tested functional on GCP and OpenStack. AWS and Azure do not provide
    an official CentOS 8 image with cloud-init support at the moment of this release.
  • Add support for Compute Canada Arbutus Cloud NVIDIA VGPUs (flavor vgpu-...).

Changed

  • Improved main documentation.
  • [AWS] Most resources if not all now have the name of the cluster as a prefix in their name
  • [OpenStack] Simplified volume attachment count computation
  • [puppet] Slurm plugin spank-cc-tmpfs_mounts is now installed from copr yumrepo
  • [puppet] Fixed order of slurm packages install
  • [puppet] Exec resource in charge of creating the slurm cluster in slurmdbd now returns 0 if the cluster already exists
  • [puppet] consul-template class initialization is now entirely in hieradata file common.yaml.
  • [puppet] CentOS 8 support: replaced notification of nfs-idmap.service by notification of nfs-server.service.
  • [puppet] CentOS 8 support: replaced pdsh by clustershell
  • [puppet] CentOS 8 support: rpc_nfs_args is now only defined if os is CentOS 7.
  • [puppet] CentOS 8 support: ipa_create_user.py now use /usr/libexec/platform-python instead of /usr/bin/env python.
  • [puppet] CentOS 8 support: Replaced Python 2 unicode calls in ipa_create_user.py by six's text_type
  • [puppet] CentOS 8 support: Moved list of nvidia package names from class profile::gpu to hieradata. List now depends on CentOS version.
  • [puppet] CentOS 8 support: Moved FreeIPA regen_cert_cmd value to hieradata. Command now depends on CentOS version.
  • [puppet] Bumped puppet-jupyterhub version to 3.3.2
  • [puppet] Update nvidia driver fact to make sure at most one version is in the output
  • [puppet] Changed logic of nvidia_grid_vgpu fact to just check if the instance flavor includes vgpu in its name
  • [puppet] CentOS 8 support: Moved default loaded CVMFS modules to hieradata. Module list now depends on CentOS version
  • [puppet] CentOS 8 support: Fixed nfs clean rbind execstop warning
  • [puppet] Replaced tcp_con_validator to check if slurmdbd is running by a wait_for ressource on slurmdbd.log regex
  • [puppet] CentOS 8 support: Fixed package name in nvidia-driver-version fact.
  • [cloud-init] Replaced reboot -n in runcmd by power_state with reboot now. This makes sure final stage of cloud-init is applied before reboot.
  • [gcp] CentOS 8 support: rewrote install_cloudinit.sh to avoid network issue at boot and install cloud-init only for the time needed. (issue #85)

Added

  • [puppet] Added support for CentOS 8 when selecting Slurm yumrepo
  • [puppet] Slurm 20 support: Added slurm_version variable to hieradata. It can be either 19 or 20.
  • [puppet] Slurm 20 support: Added PlugStackConfig parameter to slurm.conf
  • [puppet] Added slurm-perlapi package to profile::base::slurm
  • [puppet] Added exec to initialize cvmfs default.local with consul-template.
  • [puppet] Added a default node1 in slurm.conf when no slurmd has been registered yet in consul
  • [puppet] Added a require on Epel yumrepo for package fail2ban-server
  • [puppet] Added class profile::fail2ban::install
  • [puppet] CentOS 8 support: Added dependency on puppet-epel to install epel yumrepo
  • [puppet] CentOS 8 support: Enabled powertools repo
  • [puppet] CentOS 8 support: Enabled idm:DL1 stream
  • [puppet] CentOS 8 support: Added network-scripts package when os is CentOS 8
  • [puppet] CentOS 8 support: Added munge_socket selinux policy to allow confined user to submit jobs
  • [puppet] Added class profile::gpu::install
  • [puppet] Added a requirement on epel yumrepo for singularity package.
  • [puppet] Added a requirement for slurm exec create_account on slurm exec add_cluster
  • [puppet] CentOS 8 support: added class profile::mail::server
  • [puppet] Added a requirement on yumrepo epel to class jupyterhub in profile::jupyterhub::hub
  • [puppet] Added support for Compute Canada Arbutus Cloud VGPUs