Closed Bug 695960 Opened 13 years ago Closed 13 years ago

browserid: deploy to production

Categories

(Cloud Services :: Operations: Miscellaneous, task)

x86_64
Linux
task
Not set
normal

Tracking

(Not tracked)

VERIFIED FIXED

People

(Reporter: petef, Assigned: petef)

References

Details

(Whiteboard: [qa+])

- deploy an infra-sec reviewed build to prod boxes
- run base-line loadtests
- final QA pass
- browserid.org dns re-point, data migration
- whiskey
Whiteboard: [qa+]
* lower TTL on browserid.org records to 30 seconds (currently 300)
* wait 300 seconds after change pushes
* put existing browserid MySQL in read-only mode (set global read_only=true
** how does the UI handle this if we try to write?
** maybe we are better off putting up a "sorry, we're migrating" page
** /verify and js bundle serving should still work during this
* import database into new prod environment
* verify replication/data in new environment
* switch DNS to new environment, everything becomes read-write, people win
** people using lame TTL-disobeying dns servers will hit the old r/o env
* bump TTL back up to normal
* drink excellent scotch

of course before we actually do this, we should do a dry-run and import
current prod data, then have jbonacci QA with a hacked /etc/hosts. I think
we can do this early next week, after we loadtest the prod cluster.
Assignee: nobody → petef
Status: NEW → ASSIGNED
deployed!  72 second outage.
Status: ASSIGNED → RESOLVED
Closed: 13 years ago
Resolution: --- → FIXED
Done and running.
Verified by jrgm and myself.
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.