Forking from T296452#8653039

during the recent DC switch over Netbox got moved to codfw and it was super slow. this means that in the current set up:

• active/active may not be the best idea
• we need to update the dc-switch cookbook to also failover the postgresdb

The issue comes from the extra latency of having the frontend in codfw and the DB in eqiad.

There are 2 main ways of solving the issue.

• We always move the primary DB to where the primary frontend is
  • But this prevents doing active/active
• We split reads and writes, reads are always done on the local node, and writes are done where the primary DB is
  • This means slower writes, but as Netbox is read heavy this is most likely fine
  • This permits active/active (and ensure all nodes are healthy)

Option 2 seems better to me, but doesn't have any builtin support in Netbox, I've been pointed to this django module: https://github.com/jbalogh/django-multidb-router but it haven't been updated since a while.
A cookbook to ease master switchover would be valuable in all cases.