lae.proxmox

Installs and configures Proxmox Virtual Environment 6.x/7.x/8.x on Debian servers.

This role allows you to deploy and manage single-node PVE installations and PVE clusters (3+ nodes) on Debian Buster (10) and Bullseye (11). You are able to configure the following with the assistance of this role:

• PVE RBAC definitions (roles, groups, users, and access control lists)
• PVE Storage definitions
• datacenter.cfg
• HTTPS certificates for the Proxmox Web GUI (BYO)
• PVE repository selection (e.g. pve-no-subscription or pve-enterprise)
• Watchdog modules (IPMI and NMI) with applicable pve-ha-manager config
• ZFS module setup and ZED notification email

With clustering enabled, this role does (or allows you to do) the following:

• Ensure all hosts can connect to one another as root over SSH
• Initialize a new PVE cluster (or possibly adopt an existing one)
• Create or add new nodes to a PVE cluster
• Setup Ceph on a PVE cluster
• Create and manage high availability groups

Support/Contributing

For support or if you'd like to contribute to this role but want guidance, feel free to join this Discord server: https://discord.gg/cjqr6Fg. Please note, this is an temporary invite, so you'll need to wait for @lae to assign you a role, otherwise Discord will remove you from the server when you logout.

Quickstart

The primary goal for this role is to configure and manage a Proxmox VE cluster (see example playbook), however this role can be used to quickly install single node Proxmox servers.

I'm assuming you already have Ansible installed. You will need to use an external machine to the one you're installing Proxmox on (primarily because of the reboot in the middle of the installation, though I may handle this somewhat differently for this use case later).

Copy the following playbook to a file like install_proxmox.yml:

- hosts: all
  become: True
  roles:
    - role: geerlingguy.ntp
      vars:
        ntp_manage_config: true
        ntp_servers:
          - clock.sjc.he.net,
          - clock.fmt.he.net,
          - clock.nyc.he.net
    - role: lae.proxmox
      vars:
        - pve_group: all
        - pve_reboot_on_kernel_update: true

Install this role and a role for configuring NTP:

ansible-galaxy install lae.proxmox geerlingguy.ntp

Now you can perform the installation:

ansible-playbook install_proxmox.yml -i $SSH_HOST_FQDN, -u $SSH_USER

If your SSH_USER has a sudo password, pass the -K flag to the above command. If you also authenticate to the host via password instead of pubkey auth, pass the -k flag (make sure you have sshpass installed as well). You can set those variables prior to running the command or just replace them. Do note the comma is important, as a list is expected (otherwise it'll attempt to look up a file containing a list of hosts).

Once complete, you should be able to access your Proxmox VE instance at https://$SSH_HOST_FQDN:8006.

Deploying a fully-featured PVE 8.x cluster

Create a new playbook directory. We call ours lab-cluster. Our playbook will eventually look like this, but yours does not have to follow all of the steps:

lab-cluster/
├── files
│   └── pve01
│       ├── lab-node01.local.key
│       ├── lab-node01.local.pem
│       ├── lab-node02.local.key
│       ├── lab-node02.local.pem
│       ├── lab-node03.local.key
│       └── lab-node03.local.pem
├── group_vars
│   ├── all
│   └── pve01
├── inventory
├── roles
│   └── requirements.yml
├── site.yml
└── templates
    └── interfaces-pve01.j2

6 directories, 12 files

First thing you may note is that we have a bunch of .key and .pem files. These are private keys and SSL certificates that this role will use to configure the web interface for Proxmox across all the nodes. These aren't necessary, however, if you want to keep using the signed certificates by the CA that Proxmox sets up internally. You may typically use Ansible Vault to encrypt the private keys, e.g.:

ansible-vault encrypt files/pve01/*.key

This would then require you to pass the Vault password when running the playbook.

Let's first specify our cluster hosts. Our inventory file may look like this:

[pve01]
lab-node01.local
lab-node02.local
lab-node03.local

You could have multiple clusters, so it's a good idea to have one group for each cluster. Now, let's specify our role requirements in roles/requirements.yml:

---
- src: geerlingguy.ntp
- src: lae.proxmox

We need an NTP role to configure NTP, so we're using Jeff Geerling's role to do so. You wouldn't need it if you already have NTP configured or have a different method for configuring NTP.

Now, let's specify some group variables. First off, let's create group_vars/all for setting NTP-related variables:

---
ntp_manage_config: true
ntp_servers:
  - lab-ntp01.local iburst
  - lab-ntp02.local iburst

Of course, replace those NTP servers with ones you prefer.

Now for the flesh of your playbook, pve01's group variables. Create a file group_vars/pve01, add the following, and modify accordingly for your environment.

---
pve_group: pve01
pve_watchdog: ipmi
pve_ssl_private_key: "{{ lookup('file', pve_group + '/' + inventory_hostname + '.key') }}"
pve_ssl_certificate: "{{ lookup('file', pve_group + '/' + inventory_hostname + '.pem') }}"
pve_cluster_enabled: yes
pve_groups:
  - name: ops
    comment: Operations Team
pve_users:
  - name: admin1@pam
    email: [email protected]
    firstname: Admin
    lastname: User 1
    groups: [ "ops" ]
  - name: admin2@pam
    email: [email protected]
    firstname: Admin
    lastname: User 2
    groups: [ "ops" ]
pve_acls:
  - path: /
    roles: [ "Administrator" ]
    groups: [ "ops" ]
pve_storages:
  - name: localdir
    type: dir
    content: [ "images", "iso", "backup" ]
    path: /plop
    maxfiles: 4
pve_ssh_port: 22

interfaces_template: "interfaces-{{ pve_group }}.j2"

pve_group is set to the group name of our cluster, pve01 - it will be used for the purposes of ensuring all hosts within that group can connect to each other and are clustered together. Note that the PVE cluster name will be set to this group name as well, unless otherwise specified by pve_cluster_clustername. Leaving this undefined will default to proxmox.

pve_watchdog here enables IPMI watchdog support and configures PVE's HA manager to use it. Leave this undefined if you don't want to configure it.

pve_ssl_private_key and pve_ssl_certificate point to the SSL certificates for pvecluster. Here, a file lookup is used to read the contents of a file in the playbook, e.g. files/pve01/lab-node01.key. You could possibly just use host variables instead of files, if you prefer.

pve_cluster_enabled enables the role to perform all cluster management tasks. This includes creating a cluster if it doesn't exist, or adding nodes to the existing cluster. There are checks to make sure you're not mixing nodes that are already in existing clusters with different names.

pve_groups, pve_users, and pve_acls authorizes some local UNIX users (they must already exist) to access PVE and gives them the Administrator role as part of the ops group. Read the User and ACL Management section for more info.

pve_storages allows to create different types of storage and configure them. The backend needs to be supported by Proxmox. Read the Storage Management section for more info.

pve_ssh_port allows you to change the SSH port. If your SSH is listening on a port other than the default 22, please set this variable. If a new node is joining the cluster, the PVE cluster needs to communicate once via SSH.

pve_manage_ssh (default true) allows you to disable any changes this module would make to your SSH server config. This is useful if you use another role to manage your SSH server. Note that setting this to false is not officially supported, you're on your own to replicate the changes normally made in ssh_cluster_config.yml and pve_add_node.yml.

interfaces_template is set to the path of a template we'll use for configuring the network on these Debian machines. This is only necessary if you want to manage networking from Ansible rather than manually or via each host in PVE. You should probably be familiar with Ansible prior to doing this, as your method may involve setting host variables for the IP addresses for each host, etc.

Let's get that interface template out of the way. Feel free to skip this file (and leave it undefined in group_vars/pve01) otherwise. Here's one that I use:

# {{ ansible_managed }}
auto lo
iface lo inet loopback

allow-hotplug enp2s0f0
iface enp2s0f0 inet manual

auto vmbr0
iface vmbr0 inet static
    address {{ lookup('dig', ansible_fqdn) }}
    gateway 10.4.0.1
    netmask 255.255.255.0
    bridge_ports enp2s0f0
    bridge_stp off
    bridge_fd 0

allow-hotplug enp2s0f1
auto enp2s0f1
iface enp2s0f1 inet static
    address {{ lookup('dig', ansible_hostname + "-clusternet.local") }}
    netmask 255.255.255.0

You might not be familiar with the dig lookup, but basically here we're doing an A record lookup for each machine (e.g. lab-node01.local) for the first interface (and configuring it as a bridge we'll use for VM interfaces), and then another slightly modified lookup for the "clustering" network we might use for Ceph ("lab-node01-clusternet.local"). Of course, yours may look completely different, especially if you're using bonding, three different networks for management/corosync, storage and VM traffic, etc.

Finally, let's write our playbook. site.yml will look something like this:

---
- hosts: all
  become: True
  roles:
    - geerlingguy.ntp

# Leave this out if you're not modifying networking through Ansible
- hosts: pve01
  become: True
  serial: 1
  tasks:
    - name: Install bridge-utils
      apt:
        name: bridge-utils

    - name: Configure /etc/network/interfaces
      template:
        src: "{{ interfaces_template }}"
        dest: /etc/network/interfaces
      register: _configure_interfaces

    - block:
      - name: Reboot for networking changes
        shell: "sleep 5 && shutdown -r now 'Networking changes found, rebooting'"
        async: 1
        poll: 0

      - name: Wait for server to come back online
        wait_for_connection:
          delay: 15
      when: _configure_interfaces is changed

- hosts: pve01
  become: True
  roles:
    - lae.proxmox

Basically, we run the NTP role across all hosts (you might want to add some non-Proxmox machines), configure networking on pve01 with our separate cluster network and bridge layout, reboot to make those changes take effect, and then run this Proxmox role against the hosts to setup a cluster.

At this point, our playbook is ready and we can run the playbook.

Ensure that roles and dependencies are installed:

ansible-galaxy install -r roles/requirements.yml --force
pip install jmespath dnspython

jmespath is required for some of the tasks involving clustering. dnspython is only required if you're using a dig lookup, which you probably won't be if you skipped configuring networking. We pass --force to ansible-galaxy here so that roles are updated to their latest versions if already installed.

Now run the playbook:

ansible-playbook -i inventory site.yml -e '{"pve_reboot_on_kernel_update": true}'

The -e '{"pve_reboot_on_kernel_update": true}' should mainly be run the first time you do the Proxmox cluster setup, as it'll reboot the server to boot into a PVE kernel. Subsequent runs should leave this out, as you want to sequentially reboot servers after the cluster is running.

To specify a particular user, use -u root (replacing root), and if you need to provide passwords, use -k for SSH password and/or -K for sudo password. For example:

ansible-playbook -i inventory site.yml -K -u admin1

This will ask for a sudo password, then login to the admin1 user (using public key auth - add -k for pw) and run the playbook.

That's it! You should now have a fully deployed Proxmox cluster. You may want to create Ceph storage on it afterwards (see Ceph for more info) and other tasks possibly, but the hard part is mostly complete.

Example Playbook

This will configure hosts in the group pve01 as one cluster, as well as reboot the machines should the kernel have been updated. (Only recommended to set this flag during installation - reboots during operation should occur serially during a maintenance period.) It will also enable the IPMI watchdog.

- hosts: pve01
  become: True
  roles:
    - role: geerlingguy.ntp
        ntp_manage_config: true
        ntp_servers:
          - clock.sjc.he.net,
          - clock.fmt.he.net,
          - clock.nyc.he.net
    - role: lae.proxmox
        pve_group: pve01
        pve_cluster_enabled: yes
        pve_reboot_on_kernel_update: true
        pve_watchdog: ipmi

Role Variables

[variable]: [default] #[description/purpose]
pve_group: proxmox # host group that contains the Proxmox hosts to be clustered together
pve_repository_line: "deb http://download.proxmox.com/debian/pve bullseye pve-no-subscription" # apt-repository configuration - change to enterprise if needed (although TODO further configuration may be needed)
pve_remove_subscription_warning: true # patches the subscription warning messages in proxmox if you are using the community edition
pve_extra_packages: [] # Any extra packages you may want to install, e.g. ngrep
pve_run_system_upgrades: false # Let role perform system upgrades
pve_run_proxmox_upgrades: true # Let role perform Proxmox VE upgrades
pve_check_for_kernel_update: true # Runs a script on the host to check kernel versions
pve_reboot_on_kernel_update: false # If set to true, will automatically reboot the machine on kernel updates
pve_reboot_on_kernel_update_delay: 60 # Number of seconds to wait before and after a reboot process to proceed with next task in cluster mode
pve_remove_old_kernels: true # Currently removes kernel from main Debian repository
pve_pcie_passthrough_enabled: false # Set this to true to enable PCIe passthrough.
pve_iommu_passthrough_mode: false # Set this to true to allow VMs to bypass the DMA translation. This might increase performance for IOMMU passthrough.
pve_iommu_unsafe_interrupts: false # Set this to true if your system doesn't support interrupt remapping.
pve_mediated_devices_enabled: false # Set this to true if your device supports gtv-g and you wish to enable split functionality.
pve_pcie_ovmf_enabled: false # Set this to true to enable GPU OVMF PCI passthrough.
pve_pci_device_ids: [] # List of pci device ID's (see https://pve.proxmox.com/wiki/Pci_passthrough#GPU_Passthrough).
pve_vfio_blacklist_drivers: [] # List of device drivers to blacklist from the Proxmox host (see https://pve.proxmox.com/wiki/PCI(e)_Passthrough).
pve_pcie_ignore_msrs: false # Set this to true if passing through to Windows machine to prevent VM crashing.
pve_pcie_report_msrs: true # Set this to false to prevent dmesg system from logging msrs crash reports.
pve_watchdog: none # Set this to "ipmi" if you want to configure a hardware watchdog. Proxmox uses a software watchdog (nmi_watchdog) by default.
pve_watchdog_ipmi_action: power_cycle # Can be one of "reset", "power_cycle", and "power_off".
pve_watchdog_ipmi_timeout: 10 # Number of seconds the watchdog should wait
pve_zfs_enabled: no # Specifies whether or not to install and configure ZFS packages
# pve_zfs_options: "" # modprobe parameters to pass to zfs module on boot/modprobe
# pve_zfs_zed_email: "" # Should be set to an email to receive ZFS notifications
pve_zfs_create_volumes: [] # List of ZFS Volumes to create (to use as PVE Storages). See section on Storage Management.
pve_ceph_enabled: false # Specifies wheter or not to install and configure Ceph packages. See below for an example configuration.
pve_ceph_repository_line: "deb http://download.proxmox.com/debian/ceph-pacific bullseye main" # apt-repository configuration. Will be automatically set for 6.x and 7.x (Further information: https://pve.proxmox.com/wiki/Package_Repositories)
pve_ceph_network: "{{ (ansible_default_ipv4.network +'/'+ ansible_default_ipv4.netmask) | ansible.utils.ipaddr('net') }}" # Ceph public network
# pve_ceph_cluster_network: "" # Optional, if the ceph cluster network is different from the public network (see https://pve.proxmox.com/pve-docs/chapter-pveceph.html#pve_ceph_install_wizard)
pve_ceph_nodes: "{{ pve_group }}" # Host group containing all Ceph nodes
pve_ceph_mon_group: "{{ pve_group }}" # Host group containing all Ceph monitor hosts
pve_ceph_mgr_group: "{{ pve_ceph_mon_group }}" # Host group containing all Ceph manager hosts
pve_ceph_mds_group: "{{ pve_group }}" # Host group containing all Ceph metadata server hosts
pve_ceph_osds: [] # List of OSD disks
pve_ceph_pools: [] # List of pools to create
pve_ceph_fs: [] # List of CephFS filesystems to create
pve_ceph_crush_rules: [] # List of CRUSH rules to create
# pve_ssl_private_key: "" # Should be set to the contents of the private key to use for HTTPS
# pve_ssl_certificate: "" # Should be set to the contents of the certificate to use for HTTPS
pve_roles: [] # Added more roles with specific privileges. See section on User Management.
pve_groups: [] # List of group definitions to manage in PVE. See section on User Management.
pve_users: [] # List of user definitions to manage in PVE. See section on User Management.
pve_storages: [] # List of storages to manage in PVE. See section on Storage Management.
pve_datacenter_cfg: {} # Dictionary to configure the PVE datacenter.cfg config file.
pve_domains_cfg: [] # List of realms to use as authentication sources in the PVE domains.cfg config file.
pve_no_log: false # Set this to true in production to prevent leaking of storage credentials in run logs. (may be used in other tasks in the future)

To enable clustering with this role, configure the following variables appropriately:

pve_cluster_enabled: no # Set this to yes to configure hosts to be clustered together
pve_cluster_clustername: "{{ pve_group }}" # Should be set to the name of the PVE cluster
pve_manage_hosts_enabled : yes # Set this to no to NOT configure hosts file (case of using vpn and hosts file is already configured)

The following variables are used to provide networking information to corosync. These are known as ring0_addr/ring1_addr or link0_addr/link1_addr, depending on PVE version. They should be IPv4 or IPv6 addresses. You can also configure the priority of these interfaces to hint to corosync which interface should handle cluster traffic (lower numbers indicate higher priority). For more information, refer to the Cluster Manager chapter in the PVE Documentation.

# pve_cluster_addr0: "{{ defaults to the default interface ipv4 or ipv6 if detected }}"
# pve_cluster_addr1: "another interface's IP address or hostname"
# pve_cluster_addr0_priority: 255
# pve_cluster_addr1_priority: 0

You can set options in the datacenter.cfg configuration file:

pve_datacenter_cfg:
  keyboard: en-us

You can also configure HA manager groups:

pve_cluster_ha_groups: [] # List of HA groups to create in PVE.

This example creates a group "lab_node01" for resources assigned to the lab-node01 host:

pve_cluster_ha_groups:
  - name: lab_node01
    comment: "My HA group"
    nodes: "lab-node01"
    nofailback: 0
    restricted: 0

All configuration options supported in the datacenter.cfg file are documented in the Proxmox manual datacenter.cfg section.

In order for live reloading of network interfaces to work via the PVE web UI, you need to install the ifupdown2 package. Note that this will remove ifupdown. You can specify this using the pve_extra_packages role variable.

You can set realms / domains as authentication sources in the domains.cfg configuration file. If this file is not present, only the Linux PAM and Proxmox VE authentication server realms are available. Supported types are pam, pve, ad and ldap. It’s possible to automatically sync users and groups for LDAP-based realms (LDAP & Microsoft Active Directory) with sync: true. One realm should have the default: 1 property to mark it as the default:

pve_domains_cfg:
  - name: pam
    type: pam
    attributes:
      comment: Linux PAM standard authentication
  - name: pve
    type: pve
    attributes:
      comment: Proxmox VE authentication server
  - name: ad
    type: ad
    attributes:
      comment: Active Directory authentication
      domain: yourdomain.com
      server1: dc01.yourdomain.com
      default: 1
      secure: 1
      server2: dc02.yourdomain.com
  - name: ldap
    type: ldap
    sync: true
    attributes:
      comment: LDAP authentication
      base_dn: CN=Users,dc=yourdomain,dc=com
      bind_dn: "uid=svc-reader,CN=Users,dc=yourdomain,dc=com"
      bind_password: "{{ secret_ldap_svc_reader_password }}"
      server1: ldap1.yourdomain.com
      user_attr: uid
      secure: 1
      server2: ldap2.yourdomain.com

Dependencies

This role does not install NTP, so you should configure NTP yourself, e.g. with the geerlingguy.ntp role as shown in the example playbook.

When clustering is enabled, this role makes use of the json_query filter, which requires that the jmespath library be installed on your control host. You can either pip install jmespath or install it via your distribution's package manager, e.g. apt-get install python-jmespath.

User and ACL Management

You can use this role to manage users and groups within Proxmox VE (both in single server deployments and cluster deployments). Here are some examples.

pve_groups:
  - name: Admins
    comment: Administrators of this PVE cluster
  - name: api_users
  - name: test_users
pve_users:
  - name: root@pam
    email: [email protected]
  - name: lae@pam
    email: [email protected]
    firstname: Musee
    lastname: Ullah
    groups: [ "Admins" ]
  - name: pveapi@pve
    password: "Proxmox789"
    groups:
      - api_users
  - name: testapi@pve
    password: "Test456"
    enable: no
    groups:
      - api_users
      - test_users
  - name: tempuser@pam
    expire: 1514793600
    groups: [ "test_users" ]
    comment: "Temporary user set to expire on 2018年  1月  1日 月曜日 00:00:00 PST"
    email: [email protected]
    firstname: Test
    lastname: User

Refer to library/proxmox_user.py link and library/proxmox_group.py link for module documentation.

For managing roles and ACLs, a similar module is employed, but the main difference is that most of the parameters only accept lists (subject to change):

pve_roles:
  - name: Monitoring
    privileges:
      - "Sys.Modify"
      - "Sys.Audit"
      - "Datastore.Audit"
      - "VM.Monitor"
      - "VM.Audit"
pve_acls:
  - path: /
    roles: [ "Administrator" ]
    groups: [ "Admins" ]
  - path: /pools/testpool
    roles: [ "PVEAdmin" ]
    users:
      - pveapi@pve
    groups:
      - test_users

Refer to library/proxmox_role.py link and library/proxmox_acl.py link for module documentation.

Storage Management

You can use this role to manage storage within Proxmox VE (both in single server deployments and cluster deployments). For now, the only supported types are dir, rbd, nfs, cephfs, lvm,lvmthin, zfspool, btrfs, cifs and pbs. Here are some examples.

pve_storages:
  - name: dir1
    type: dir
    content: [ "images", "iso", "backup" ]
    path: /ploup
    disable: no
    maxfiles: 4
  - name: ceph1
    type: rbd
    content: [ "images", "rootdir" ]
    nodes: [ "lab-node01.local", "lab-node02.local" ]
    username: admin
    pool: rbd
    krbd: yes
    monhost:
      - 10.0.0.1
      - 10.0.0.2
      - 10.0.0.3
  - name: nfs1
    type: nfs
    content: [ "images", "iso" ]
    server: 192.168.122.2
    export: /data
  - name: lvm1
    type: lvm
    content: [ "images", "rootdir" ]
    vgname: vg1
  - name: lvmthin1
    type: lvmthin
    content: [ "images", "rootdir" ]
    vgname: vg2
    thinpool: data
  - name: cephfs1
    type: cephfs
    content: [ "snippets", "vztmpl", "iso" ]
    nodes: [ "lab-node01.local", "lab-node02.local" ]
    monhost:
      - 10.0.0.1
      - 10.0.0.2
      - 10.0.0.3
  - name: pbs1
    type: pbs
    content: [ "backup" ]
    server: 192.168.122.2
    username: user@pbs
    password: PBSPassword1
    datastore: main
    namespace: Top/something # Optional
  - name: zfs1
    type: zfspool
    content: [ "images", "rootdir" ]
    pool: rpool/data
    sparse: true
  - name: btrfs1
    type: btrfs
    content: [ "images", "rootdir" ]
    nodes: [ "lab-node01.local", "lab-node02.local" ]
    path: /mnt/proxmox_storage
    is_mountpoint: true
  - name: cifs1
    server: cifs-host.domain.tld
    type: cifs
    content: [ "snippets", "vztmpl", "iso" ]
    share: sharename
    subdir: /subdir
    username: user
    password: supersecurepass
    domain: addomain.tld

Refer to https://pve.proxmox.com/pve-docs/api-viewer/index.html for more information.

Currently the zfspool type can be used only for images and rootdir contents. If you want to store the other content types on a ZFS volume, you need to specify them with type dir, path /<POOL>/<VOLUME> and add an entry in pve_zfs_create_volumes. This example adds a iso storage on a ZFS pool:

pve_zfs_create_volumes:
  - rpool/iso
pve_storages:
  - name: iso
    type: dir
    path: /rpool/iso
    content: [ "iso" ]

Refer to library/proxmox_storage.py link for module documentation.

Ceph configuration

This section could use a little more love. If you are actively using this role to manage your PVE Ceph cluster, please feel free to flesh this section more thoroughly and open a pull request! See issue #68.

PVE Ceph management with this role is experimental. While users have successfully used this role to deploy PVE Ceph, it is not fully tested in CI (due to a lack of usable block devices to use as OSDs in Travis CI). Please deploy a test environment with your configuration first prior to prod, and report any issues if you run into any.

This role can configure the Ceph storage system on your Proxmox hosts. The following definitions show some of the configurations that are possible.

pve_ceph_enabled: true
pve_ceph_network: '172.10.0.0/24'
pve_ceph_cluster_network: '172.10.1.0/24'
pve_ceph_nodes: "ceph_nodes"
pve_ceph_osds:
  # OSD with everything on the same device
  - device: /dev/sdc
  # OSD with block.db/WAL on another device
  - device: /dev/sdd
    block.db: /dev/sdb1
  # encrypted OSD with everything on the same device
  - device: /dev/sdc
    encrypted: true
  # encrypted OSD with block.db/WAL on another device
  - device: /dev/sdd
    block.db: /dev/sdb1
    encrypted: true
# Crush rules for different storage classes
# By default 'type' is set to host, you can find valid types at
# (https://docs.ceph.com/en/latest/rados/operations/crush-map/)
# listed under 'TYPES AND BUCKETS'
pve_ceph_crush_rules:
  - name: replicated_rule
    type: osd # This is an example of how you can override a pre-existing rule
  - name: ssd
    class: ssd
    type: osd
    min-size: 2
    max-size: 8
  - name: hdd
    class: hdd
    type: host
# 2 Ceph pools for VM disks which will also be defined as Proxmox storages
# Using different CRUSH rules
pve_ceph_pools:
  - name: ssd
    pgs: 128
    rule: ssd
    application: rbd
    storage: true
# This Ceph pool uses custom size/replication values
  - name: hdd
    pgs: 32
    rule: hdd
    application: rbd
    storage: true
    size: 2
    min-size: 1
# This Ceph pool uses custom autoscale mode : "off" | "on" | "warn"> (default = "warn")
  - name: vm-storage
    pgs: 128
    rule: replicated_rule
    application: rbd
    autoscale_mode: "on"
    storage: true
pve_ceph_fs:
# A CephFS filesystem not defined as a Proxmox storage
  - name: backup
    pgs: 64
    rule: hdd
    storage: false
    mountpoint: /srv/proxmox/backup

pve_ceph_network by default uses the ansible.utils.ipaddr filter, which requires the netaddr library to be installed and usable by your Ansible controller.

pve_ceph_nodes by default uses pve_group, this parameter allows to specify on which nodes install Ceph (e.g. if you don't want to install Ceph on all your nodes).

pve_ceph_osds by default creates unencrypted ceph volumes. To use encrypted volumes the parameter encrypted has to be set per drive to true.

PCIe Passthrough

This role can be configured to allow PCI device passthrough from the Proxmox host to VMs. This feature is not enabled by default since not all motherboards and CPUs support this feature. To enable passthrough, the devices CPU must support hardware virtualization (VT-d for Intel based systems and AMD-V for AMD based systems). Refer to the manuals of all components to determine whether this feature is supported or not. Naming conventions of will vary, but is usually referred to as IOMMU, VT-d, or AMD-V.

By enabling this feature, dedicated devices (such as a GPU or USB devices) can be passed through to the VMs. Along with dedicated devices, various integrated devices such as Intel or AMD's integrated GPU's are also able to be passed through to VMs.

Some devices are able to take advantage of Mediated usage. Mediated devices are able to be passed through to multiple VMs to share resources, while still remaining usable by the host system. Splitting of devices is not always supported and should be validated before being enabled to prevent errors. Refer to the manual of the device you want to pass through to determine whether the device is capable of mediated usage (Currently this role only supports GVT-g; SR-IOV is not currently supported and must be enable manually after role completion).

The following is an example configuration which enables PCIe passthrough:

pve_pcie_passthrough_enabled: true
pve_iommu_passthrough_mode: true
pve_iommu_unsafe_interrupts: false
pve_mediated_devices_enabled: false
pve_pcie_ovmf_enabled: false
pve_pci_device_ids:
  - id: "10de:1381"
  - id: "10de:0fbc"
pve_vfio_blacklist_drivers:
  - name: "radeon"
  - name: "nouveau"
  - name: "nvidia"
pve_pcie_ignore_msrs: false
pve_pcie_report_msrs: true

pve_pcie_passthrough_enabled is required to use any PCIe passthrough functionality. Without this enabled, all other PCIe related fields will be unused.

pve_iommu_passthrough_mode enabling IOMMU passthrough mode might increase device performance. By enabling this feature, it allows VMs to bypass the default DMA translation which would normally be performed by the hyper-visor. Instead, VMs pass DMA requests directly to the hardware IOMMU.

pve_iommu_unsafe_interrupts is required to be enabled to allow PCI passthrough if your system doesn't support interrupt remapping. You can find check whether the device supports interrupt remapping by using dmesg | grep 'remapping'. If you see one of the following lines:

• "AMD-Vi: Interrupt remapping enabled"
• "DMAR-IR: Enabled IRQ remapping in x2apic mode" ('x2apic' can be different on old CPUs, but should still work)

Then system interrupt remapping is supported and you do not need to enable unsafe interrupts. Be aware that by enabling this value your system can become unstable.

pve_mediated_devices_enabled enables GVT-g support for integrated devices such as Intel iGPU's. Not all devices support GVT-g so it is recommended to check with your specific device beforehand to ensure it is allowed.

pve_pcie_ovmf_enabled enables GPU OVMF PCI passthrough. When using OVMF you should select 'OVMF' as the BIOS option for the VM instead of 'SeaBIOS' within Proxmox. This setting will try to opt-out devices from VGA arbitration if possible.

pve_pci_device_ids is a list of device and vendor ids that is wished to be passed through to VMs from the host. See the section 'GPU Passthrough' on the Proxmox WIKI to find your specific device and vendor id's. When setting this value, it is required to specify an 'id' for each new element in the array.

pve_vfio_blacklist_drivers is a list of drivers to be excluded/blacklisted from the host. This is required when passing through a PCI device to prevent the host from using the device before it can be assigned to a VM. When setting this value, it is required to specify a 'name' for each new element in the array.

pve_pcie_ignore_msrs prevents some Windows applications like GeForce Experience, Passmark Performance Test and SiSoftware Sandra from crashing the VM. This value is only required when passing PCI devices to Windows based systems.

pve_pcie_report_msrs can be used to enable or disable logging messages of msrs warnings. If you see a lot of warning messages in your 'dmesg' system log, this value can be used to silence msrs warnings.

Developer Notes

When developing new features or fixing something in this role, you can test out your changes by using Vagrant (only libvirt is supported currently). The playbook can be found in tests/vagrant (so be sure to modify group variables as needed). Be sure to test any changes on both Debian 10 and 11 (update the Vagrantfile locally to use debian/buster64) before submitting a PR.

You can also specify an apt caching proxy (e.g. apt-cacher-ng, and it must run on port 3142) with the APT_CACHE_HOST environment variable to speed up package downloads if you have one running locally in your environment. The vagrant playbook will detect whether or not the caching proxy is available and only use it if it is accessible from your network, so you could just permanently set this variable in your development environment if you prefer.

For example, you could run the following to show verbose/easier to read output, use a caching proxy, and keep the VMs running if you run into an error (so that you can troubleshoot it and/or run vagrant provision after fixing):

APT_CACHE_HOST=10.71.71.10 ANSIBLE_STDOUT_CALLBACK=debug vagrant up --no-destroy-on-error

Contributors

Musee Ullah (@lae, lae@lae.is) - Main developer
Fabien Brachere (@Fbrachere) - Storage config support
Gaudenz Steinlin (@gaundez) - Ceph support, etc
Richard Scott (@zenntrix) - Ceph support, PVE 7.x support, etc
Thoralf Rickert-Wendt (@trickert76) - PVE 6.x support, etc
Engin Dumlu (@roadrunner)
Jonas Meurer (@mejo-)
Ondrej Flidr (@SniperCZE)
niko2 (@niko2)
Christian Aublet (@caublet)
Gille Pietri (@gilou)
Michael Holasek (@mholasek)
Alexander Petermann (@lexxxel) - PVE 8.x support, etc
Bruno Travouillon (@btravouillon) - UX improvements
Tobias Negd (@wu3rstle) - Ceph support
PendaGTP (@PendaGTP) - Ceph support
John Marion (@jmariondev)
foerkede (@foerkede) - ZFS storage support
Guiffo Joel (@futuriste) - Pool configuration support
Adam Delo (@ol3d) - PCIe Passthrough Support

Full list of contributors