Impacts after migration:

• [1] still reachable as before
• the machine shall be reached at riverside.internal.admin.swh.network (ssh).

Note:
Node exposing sentry service: riverside.internal.softwareheritage.org [2].

[1] https://sentry.softwareheritage.org

[2] https://inventory.internal.admin.swh.network/virtualization/virtual-machines/12/

Step-by-step plan:

• Gandi: Reduce sentry.s.o CNAME ttl early (days before migration starts, e.g. ~300s)
• Inventory:
  • Reserve new ip in vlan 442
  • Deprecate the ip from vlan 440
• D7045: Puppet manifest adaptations for moving the node to the admin vlan [4]
• Firewall: Open rule to allow access from pergamon to riverside:9000
• On {pergamon, riverside, rp1} [5]
  • Stop puppet agent
• On pergamon
  • Deploy new puppet manifest change (last time we forgot ¯\_(ツ)_/¯)
• On riverside:
  • Update the ip to the new vlan442 ip (192.168.50.70)
    • Connect through ssh and adapt /etc/network/interfaces with new ip
    • Modify directly through the proxmox ui (not terraform-ed yet)
    • Adapt hardware entry about network (proxmox ui) to change from vmbr0 to vmbr442
  • Update the hostname to riverside.i.a.s.n
  • Remove the puppet certificates rm -rf /var/lib/puppet/ssl (agent node)
  • Update facts deployment and subnets /etc/facter/facts.d/deployment.txt to admin [6]
  • Reboot machine (poweroff, start)
  • Run puppet with puppet agent --test --fqdn riverside.internal.admin.swh.network
  • Install necessary facts for cloud-init to stop tampering with /etc/hosts
• On pergamon:
  • Run puppet agent
  • Decommission riverside.i.s.o certificate
• On rp1:
  • Run puppet agent
• Gandi: Change sentry.s.o CNAME value from pergamon to swh-rproxy3.inria.fr. (to target the admin reverse proxy)
• Inventory:
  • Change the reserved ip status to active
  • Update sentry node with its new ip [1]
• Clean up no longer necessary sentry reverse proxy on pergamon
• Gandi: Bump the sentry.s.o CNAME TTL to "standard" value of 1800 (like the others)
• Terraform:
  • Reference riverside node in sysadm terraform admin manifest [3] node is diverging too much, the risk/benefit seems off so we don't do it.

[3] https://forge.softwareheritage.org/source/swh-sysadmin-provisioning/browse/master/proxmox/terraform/admin/admin.tf

[4] Check the diff description/code for more details

[5]

$ clush -b -w pergamon -w riverside -w rp1.internal.admin.swh.network "puppet agent --disable T3891"

[6]

root@riverside:~# cat /etc/facter/facts.d/deployment.txt
deployment=admin
root@riverside:~# cat /etc/facter/facts.d/subnet.txt
subnet=sesi_rocquencourt_admin