Add a new hypervisor

Louvre is no longer able to run all existing VMs. A third hypervisor has to be added to our infrastructure in order to continue to operate smoothly when one of them is out of service.

---

Migrated from T1392 (view on Phabricator)

marked this issue as related to #1173 (closed)

added System administration priority:Normal state:wip labels

New hypervisor hardware has been racked in our bay at Rocquencourt. The machine's iDrac management interface is accessible on the management network, under the name swh7-adm.inria.fr (details on the wiki).

Proxmox now installed on the machine, hypervisor3.softwareheritage.org.

I've reinstalled the machine following these steps:

• Debian installed on the machine with the plain debian installer (no bonding support => no network)
• network configured with iproute2 (with the bond + vlan stack)
  • https://baturin.org/docs/iproute2/ was quite useful
• install facter from stretch backports, install puppet from stretch
• /etc/facter/facts.d seeded
• puppet run
• install ifenslave, bridge-utils and vlan to get the proper /etc/network/interfaces scripts
• write proper /etc/network/interfaces with inspiration from louvre
• reboot and hope the network comes up. rinse, repeat.
• setup proxmox as advised in https://pve.proxmox.com/wiki/Install_Proxmox_VE_on_Debian_Stretch
• reboot on the proxmox kernel
• Update /etc/pve/corosync.conf with the new host on one of the existing hypervisors
  • add node to nodelist with a new nodeid
  • add unicast ip address to ring0 in the totem section
  • increment config version number in the totem section
• restart pve-cluster on existing nodes to take the new corosync config into account
• follow https://pve.proxmox.com/wiki/Cluster_Manager#pvecm_join_node_to_cluster
  • pvecm add 192.168.100.1
  • type root password for louvre
• see the host appear in the proxmox cluster
• reboot one last time for good measure (and to see whether everything starts ok on boot)

marked this issue as related to #1503 (closed)

The new hypervisor has been working without any particular issue since its installation.

removed state:wip label

closed

mentioned in issue #1467 (closed)