Merge autopkgtest-cloud:open-release-deets into autopkgtest-cloud:master

[Brian Murray]

We encountered some issues when creating a new series in autopkgtest in part because some of the documentation was open to interpretation or not very detailed. These changes should add more details regarding when needs to happen and when it can happen.

[Steve Langasek (vorlon)]

review: Approve

[Brian Murray (brian-murray)]

Fix this

review: Needs Fixing

[Paride Legovini (paride)]

Looks good, but maybe clarify that `systemctl start download-all-results.service` is to be done on the machines hosting the autopkgtest-web units.

[Steve Langasek (vorlon)]

On Tue, Nov 01, 2022 at 03:54:04PM -0000, Paride Legovini wrote:
> Looks good, but maybe clarify that `systemctl start
> download-all-results.service` is to be done on the machines hosting the
> autopkgtest-web units.

The last time I did this part of the opening, I think I ran it on one web
instance and then scp'ed the result across to the other, which seems more
efficient?

[Paride Legovini (paride)]

Well that's an option, but I don't see it saving time if download-all-results is started in parallel on the two web instances (as we did). But that's something we can suggest, I'd only be slightly concerned of somehow corrupting the database and then copying it over corrupted to the other web instance.

<bikeshed> On this diff more in general: I'd make updating distro-info-data the very first step, and drop the first added paragraph ("Updating distro-info-data and building the images are not blocked ..."), to keep the instructions fully linear.

diff --git a/docs/administration.rst b/docs/administration.rst
index 997bfa1..c2dc434 100644
--- a/docs/administration.rst
+++ b/docs/administration.rst
@@ -142,24 +142,30 @@ usually indicate that the LXD host has gone down and needs redeploying.
 Opening up a new series
 -----------------------
 
+Updating distro-info-data and building the images are not blocked on test results
+being copied forward (``seed-new-release``) or devel results existing
+(``download-all-results``) i.e. do them while waiting for those.
+
 * Download the latest ``autopkgtest.db`` from the website/unit to the home
   directory on wendigo
 * Run ``autopkgtest-cloud/tools/seed-new-release <old_release> <new_release> autopkgtest.db``
   on wendigo.
+* Make sure an updated distro-info-data with the new series is available and
+  install it on all worker, web, and haproxy nodes. (If not yet available,
+  temporarily hack the new series into the ``/usr/share/distro-info/ubuntu.csv``
+  on them.)
 * Update the ``service-bundle`` to include the release in ``releases`` and
-  deploy it. Run ``systemctl start download-all-results.service`` to download
-  the results from swift to the db. TODO: This should be done automatically
-  by adding the release.
-* Make sure a new distro-info with the new series is available and install it
-  on all worker and web nodes. (If not yet available, temporarily hack in the
-  new series into the distro-info-data ubuntu.csv on all worker and web nodes.)
+  deploy it by using ``mojo run``. Run ``systemctl start
+  download-all-results.service`` (on the instances providing autopkgtest-web)
+  to download the results from swift to the db.
+  TODO: This should be done automatically by adding the release.
 * Build new lxd images on the lxd-armhf leader (see :ref:`Creating new LXD
   images before official ones are available`).
 * Build cloud images::
 
  sudo systemctl start build-adt-image@<release>-<cloud>-<arch>.service ...
 
-* Notify release team to remove cowboy disablement of proposed-migration,
+* Notify the release team to remove cowboy disablement of proposed-migration,
   and manually run ``run-proposed-migration`` as ``ubuntu-archive@snakefruit``
   to do a test run of proposed-migration.
 * Submit a test job for all arches via ``request.cgi`` or ``run-autopkgtest``