This runs a full install sequence. Ready for testing on preview.

Adds the access tables to its own startup. Admin API relies on it, or it
won't start.

How did I miss those?

We should be able to inspect more tables, for debugging.

This brings up the full API, and is ready for performance testing in
`preview`.

It is cleaner than it was, and may make it easier to debug the `preview`
deploy.

Adds grabbing sling into the mix

Testing again

Also, added partitions.

Partitions are 3x faster for downloading all the data (by year), but 10x
slower than batches (via EXPLAIN). In the real, it is 2x slower to
download than batches, but 3x faster than a straight download.

So. Probably worth it, if we can document that people downloading all
the data should either 1) do so year-by-year, or 2) use batches.

And, redeploying...

Need to just return a non-zero exit code...

Need to *always* succeed...

We don't really want this, but will it allow functions to return
non-zero?

Setting a global function response variable...

This was work-in-flight. Instead of pushing further, I'm committing the in-between state. This moved some Admin API stuff around, which I'll remove in the next commit.

The config change is for local dev only. We had a bucket called `gsa-fac-private-s3`, which was legacy naming that has been superceded in all other environments. The bucket is now called `fac-private-s3`, mirroring the production environments. This should not impact any operations except for local/Docker execution. (And, it should be transparent)

The next commit should clean this up further.

Probably wanted more incremental commits.

This shuffled the API around, and has a pytest for the API that can be
run locally against the local stack or the remote/live environments.

Resolves the conflict(s). Things that Bobby and I both touched; nothing in the mainline app.

Our local environment has to make sure that fac-snapshot-db looks
"right" before we do anything. This means making sure we run cgov-util
to do a snapshot before we do anything pre/post with SQL.

This introduces a new copy of the data, so that the tables we back up to
every night do not become an issue. (That is, we want the API bound to
the copy, not to the copies that get dumped every night.)

This lets the local config (at least for bash purposes) look almost
exactly like the cloud.gov setup.

This keeps the container from dropping products like cgov-util into our
local filesystem.

This is now "just SQL," and does not need to be managed by the
management command.

It should run in the `pre` sequence. This way, the triggers are
configured early in the process. We then run the `--disable` promptly
afterwards.

These are redundant; they're issued earlier in the sequence.

Correctly uses DB2 for advanced search.

May want to implement it differently.

Changes for this in one commit, to make finding it easy.

Admin Panel
- Introducing 2 new groups - `helpdesk` and `readonly`. These groups will be used with new permissions for performing Admin API operations directly through the UI instead.
- Introducing 2 new tables, which were previously accessible through the Admin API. Helpdesk will have CRUD operations on these tables.

StaffUser management
- Introduced a `staffusers.json` file, which will now manage the permissions for the staff in respect to the Admin UI.
- New mgmt command which is handled through `run.sh` to create the staff users according to the above JSON list.
- Updated onboarding document to reflect a new user joining the Admin Panel.

OTHER
- linting.

The move out was part of debugging, and unnecessary. This reduces the
number of files, and keeps like-with-like.

Permissions simplification/removal; picked up elsewhere/later in the
sequence. (And, the alter on sequences may not have been necessary at
all.)

…api-perf

Removes an unnecessary step.

This is a proof-of-concept included with the work. It's in the tree, and
I've taken it out of active use/an active part of startup.

However, we might consider working it into the nightly, to start testing
it.

This used to do a cgov-util copy. Now, it doesn't. The startup sequence
should take care of it.

This was out of date. Retrying preview deploy.

Ref: 8d59859

We need to always load the fake VCAP_SERVICES from the file when running
locally.

There's no reason to order the results from the mega-table.

Now, pre-post is per-db.

That means we must run the pre/post on SNAPSHOT separately in an action
before deploy.

Trying to figure out a bug in staffusers...

:w

…api-perf

- Fixed an issue where Cypress was checking an undefined method type `HEAD` for the existence of the PDF report.
- Now, we are just retrieving it with `GET` and checking for a valid status code.

1. Updates terraform with correct APIs.
2. Removes more schemas
3. Aligns compose with our changes

This makes it possible to target a DB that is tunneled, as opposed to
only using the containerized DB.

Handy for testing in the preview environment.

Uses external env vars to do this, so is safe.

:wq

Added max timeout to request

Now uses Django factory testing (like we do with our other tests that fall under `manage.py test`) for the tests that were written up by Matt.
INCOMPLETE - check the "FIXME" comment. Currently the environment is hardcoded to "local" under the `ApiTests` class, and I haven't gotten a "success" response from the following tests.
- `test_suppressed_not_accessible_with_bad_key`
- `test_suppressed_accessible_with_good_key`