Infrastructure » sepia

Week before
---
☑ get role updates merged
☑ get inventory updates merged
☑ run role
☑ verify jobs can be scheduled and run from teuthology-new (using teuthology's beanstalkd)
☑ sync homes
☑ write up plan
☑ include plan ticket in notice on status.front
☑ send email including plan

Monday morning
---
☑ pause vps queue
☑ add DNS A record to point teuthology-old at old host (leave teuthology alone for now)

Monday afternoon
---
☑ shut down miras that need to physically move
☑ no vps jobs should be running
☑ tell workers to stop
☑ kill vps workers if necessary
☑ unpause vps queue

Tuesday morning
---
☑ no workers should be running (and therefore no jobs)
☑ sync homes again
☑ stop beanstalkd on old host
☑ move beanstalkd data to new host
☑ start beanstalkd on new host
☑ move crontabs
☑ build necessary virtualenvs
☑ set timezone to UTC
☑ update DNS (and DHCP names) to point teuthology to new host
☑ start workers on new host
☑ send email, throw party