Infrastructure

Thorsten: So far no news from Tollef. Can you ping him again, you might know him a bit better? It blocks currently on his work/feedback...

We're still waiting for feedback from FreeDesktop.org
If you do not hear back within a week from me, we should go on with setting up a VM, where then Robinson and Joel can install BugZilla onto (so they need access to it as well)

The plan is:

- TDF sets up a VM and takes care of getting BugZilla per se (that is, without data so far) up and running there

- We then need at least read-only access to the underlying database and the files of the current instance (IIRC, one can attach files to BugZilla that do not end up in the database); either directly on the host so we grab it ourselves, or as an archive somewhere

- With a first copy of this data, we then try to put the data into our BugZilla, to end up with a clone that should have the same functionality as the current FDO instance

- If the data import works, we arrange a date and time where the FDO instance will be closed/put read-only (most probably only Tollef can do that), we do another database and file dump, and put our new instance live

So:
Tollef took care of creating a sanitized backup of fdo bugzilla
I downloaded it on vm14 and restored the postgress 'bugzilla' database (user bugz)
I picked the same name than were in the backup for simplicity sake

what is left:

1/ install bugzilla. note that the version schema of the bugzilla database we receive is 3.00 so we need a version of bugzilla that match that schema.
2/ configure nginx to serve bugzilla pages
3/ sanity check that the restored database is usable and contains what we expect. Note: I dids some select etc.. directly in the database, it looks quite good.
4/ drop the bugzilla database and re-restore it.. make sure bugzilla is not unhappy with that.

if these step are satisfactory:
4/ prepare a production VM
5/ install and prepare bugzilla on it...
6/ restore a test backup.. make sure everything is fine.. then drop it again

7/ prepare a landing static web-page to be displayed while the migration is on-going at bugs.libreoffice.org

8/ schedule a migration date 'D' with tollef.

9/ D - 72/48 hours... (depending on the current ttl) change the ttl on bugs.libreoffice.org to a small value... 5 minutes ?

10/ D-Day... meet Tollef on IRC.. make sure everything is ready...
11/ fdo turn their ugzilla off for maintenance
12/ we switch the DNS entry for bugs.bugzilla.org to our production box.. which display the landing page of 7/
13/ fdo does a backup, restore, scrub and rebackup (how long that will take is TDB.. but a couple of hours max is anticipated for now)
14/ we download the sanitized backup (~30 minutes)
15/ we restore the database (~10 minutes)
16/ we swithc bugs.libreoffice/org to bugzilla itself instead of the landing page
17/ Tollef put in place a set of redirect rules on fdo side to redirect individual bugs url to us, for the bugs that are LibreOffice related.
18/ fdo turn off LibreOffce project on their bugzilla, and scrub things up.
19/ restore the bug.libreoffice.org ttl to a more reasonable value...

20/ we remotely share a beer :-)

as a side note. we will then have both gerrit and bugzilla being heavy user of postgresql... it would be a good idea to look into database replication... to replicate both gerrit and bugzilla to a 3rd box.. that would provide a continuous hot backup.

Setting up planning meeting this week with Joel and Norbert.
Aiming to layout roadmap and schedule.