Upgrade Postgres on production

[KlaasH]

The second issue of the upgrade. Issue #859 should result in detailed instructions for how to upgrade production (and whether or not doing so requires a code deploy).

[BryanQuigley]

• [x] Upgrade bastion (may need sudo amazon-linux-extras install postgresql13)
• [x] Switch to t3 instance type (5 minutes)
• [x] ALTER EXTENSION postgis UPDATE TO "2.5.5"; (run twice then log back in to confirm)
• [x] Update database from 9.6.23->12.8 (<30) - use default parameter group
• [x] SELECT PostGIS_Extensions_Upgrade(); to get to 3.1.4 (run twice then log back in to confirm)
• [x] Update database from 12.8->13.6 (<20) - use default parameter group
• [x] SELECT PostGIS_Extensions_Upgrade(); to get to 3.1.5 (run twice then log back in to confirm)
• [x] \timing then VACUUM (FULL, ANALYZE, VERBOSE);
• [x] Update terraform variables to match and show no rds changes

Check commands to run before doing a long operation: SELECT FROM pg_available_extensions WHERE installed_version IS NOT NULL; SELECT FROM pg_available_extension_versions WHERE name in ('postgis', 'plpgsql'); \l+ (database sizes)

Testing VACUUM FULL went from 694 MB to 500 MB in .5 minutes.

Items to confirm before production deploy:

• Is t3.small fine or should we go with t3.medium (memory usage seems fine)?
• Drop MultiAZ to speed up upgrade process (and leave it off?)

[BryanQuigley]

Vaccum 42 GB -> 32 GB

INFO: "django_content_type": scanned 1 of 1 pages, containing 22 live rows and 0 dead rows; 22 rows in sample, 22 estimated total rows VACUUM Time: 3206160.835 ms (53:26.161)

Note to my future self - put "Items to confirm before production deploy:" before a bunch of instructions...