Obsolete:Labs Eqiad Migration/Fallback

From Wikitech
Jump to navigation Jump to search

NOTE: This ship has sailed; this page is only here for archival purposes and it will soon be deleted.

half-assed

Replicate the tampa cluster in eqiad:

Apply the virt2 puppet classes to labnet1001 [1]
Apply virt0 puppet to virt1000 [1]
Manually configure mediawiki on virt1000, copy over wiki from virt0 [1]
Apply virtx puppet to virt100x [1]
Make keystone aware of the new domain [2]
An eternity of debugging to make all of the above actually work

Incremental improvements to above:

Deploy virt1000 mediawiki using trebuchet/scap/whatever [3]
experiment with OpenStack version upgrade but only if all current services work with existing setup [2]

More difficult but future-proof alternatives

Adopt more modern, current OpenStack tech:

Switch to upstream puppet classes and deploy the latest OS version [2]
Switch our existing network strategy from nova-network to Neutron [4]
This may be required to use recent OS versions; I don't know if Havana supports our current network setup.
Switch current DNS model to 'designate' [4]
This doesn't need to block the migration, can happen later.
  1. 1.0 1.1 1.2 1.3 Andrew pretty much knows how to do these things already.
  2. 2.0 2.1 2.2 Andrew knows how this works and could implement with some tinkering.
  3. Plenty of knowledge about how to do this in the org already.
  4. 4.0 4.1 New territory, would require research and lots of experimentation