Improve telemetry with deeper insights into system usage

[licitdev]

Scope

What's changed:

• Filter by active users instead of all users
• Filter by enabled extensions instead of all extensions
• Add field counts: fields_max & fields_total
• Add database size

Potential Risks / Drawbacks

• More queries to the DB.

Review Notes / Questions

• The queries had been tested on the DB containers, but wrapped in a try-catch in case it fails.

---

Fixes #21982

[changeset-bot]

🦋 Changeset detected

Latest commit: b4645f7

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 2 packages

Name	Type
@directus/api	Patch
directus	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[paescuj]

Great work ✨

• For the newly introduced data (db size, and especially fields count) do we expect them to perform well? Wondering after seeing this comment:

  directus/api/src/telemetry/utils/get-item-count.ts

  Lines 49 to 50 in b4645f7

  	// Counts can be a little heavy if the table is very large, so we'll only ever execute 3 of these
  	// queries simultaneously to not overload the database

• Just because it might be good to know (and maybe add inline docs): Do (some of) the DB size queries require some special DB permissions?

[paescuj]

Should we return null here as well as for failing getDatabaseSize calls in the dialects implementation so we can identify / distinguish these cases?

[DanielBiegler]

Its not because of this PR but I noticed that we are actually counting bundles slightly wrong. Because a bundle is just a row and the inside-extensions point to it via its id a bundle with one extension inside counts as 2 and a bundle with two extensions inside counts as 3 - etc.

I looked up the original PR for the get-item-count.ts file, here, which already introduced this behavior but I dont see it being discussed there. Did we decide that we are OK with this somewhere else? (cc @rijkvanzanten maybe you know?)

Now that we added a possibility for a where clause we might be able to filter those out! I personally would be in favor of doing so. 😄 👍

[DanielBiegler]

Maybe theres an easier way but off the top of my head I came up with this:

SELECT COUNT(id)
FROM directus_extensions
WHERE id NOT IN (
  SELECT DISTINCT x1.id 
  FROM directus_extensions as x1
  INNER JOIN directus_extensions as x2 ON x1.id = x2.bundle
);

[rijkvanzanten]

A bundle should be counted as the sum of it's child extensions, excluding the parent 👍🏻