1. Open Problems After Migration
/etc/keytabs/user.daemon/* is still manually synced and not managed by Puppet
2. General Plan
- Debian packages updated for stretch
create proper signed repository on ServerNavajos, rebuild packages
- config-packages may be dropped here
- Board votes on new provider
- Account at provider set up
- New primary and secondary afs and kerberos servers set up
- Small 10G block storage attached to primary, expanded once we begin moving volumes. No block storage needed for secondary.
- Added to existing cluster as secondaries
- Create test volumes, check cross-site performance, make sure moving volumes works reliably
- Set up puppet [might be able to preload before new servers set up]
simple puppetdb + manual git pull + puppet apply from cron on each server
- convert config-packaged config to puppet classes as needed
- Set up new members vm
- Easy setup, allow members to test on stretch as early as possible
- Mail server set up
exim config rebased on latest, test keep_environment
- exim, courier, mailman, ejabberd
- migrate mailman (no reason to delay for other services, everything is local to the server)
- New web server set up
DomTool ported to apache 2.4 (reuse 2.2 config if needed, can deal with new Require syntax later)
- Defer on PHP 7.x until after dropping colo expense, unless it proves easy to support in parallel with 5.6
- Only support fastcgi php, suphp is 100% dead
- Databases setup
- Keep mysql 5.6, upgrade to postgres 10
- dbs are so small, probably just do a flag day and suffer a few minutes downtime
At this point, we can start moving members. Volunteers first, followed by folks using daemons on bog, then everyone.
- Volunteers manually (moving vhosts and primary mx to new servers, then migrating volumes)
- HCoop services and volumes
- Mail for all members and webmail interfaces
- Anyone needing to move daemons + other services at once from bog
Final move of all remaining volumes and changes defaults for DomTool to new servers
2.1. Secondary Goals
- Enable IPv6 for all services that can support it
- Assuming there's no/minimal risk of people with broken ipv6 being unable to access hcoop
3. Leaving Colo
- What to do with hardware?
- Need to scrub hard drives (mail to board member for manual destruction?), datacenter has on-site server recycling so we could just let it be recycled (essentially worth $0 at this point)
- Check bandwidth overage rates... may need to limit transfer rates to avoid, but weigh against another month of hosting
- Terminate contract
- have members migrated two weeks ahead of time ideally...