Feature: move documents count to analytics node (if exists)

[arekborucki]

[coyotte508]

TBH not sure - it's something very specific to our atlas deployment (unless proven otherwise)

IMO it's fine to keep current code (a small improvement would be to make 4 separate http requests maybe, and display immediately the dropdown with loading icons for each count)

But current code works well enough IMO not a priority

[arekborucki]

it is also working for normal deployments without analytic node

// Falls back to secondaryPreferred if no ANALYTICS node exists
const analyticsReadPreference = new ReadPreference("secondaryPreferred", [{ nodeType: "ANALYTICS" }, {}]);

my private cluster, simple replica set:

[coyotte508]

yes, my concern is that it's an optimization for us for a specific collection, and it's an open source project. Other users would possibly have different results, and this optimization is not relevant for them (eg if their analytics node is delayed or w/e)

Maybe an alternative is adding an env var for read preference to use by default? And use that for all the "read" queries (including where we used secondaryPreferred earlier)

That way for our case we can make mongoku use the analytics node always, by changing the env var, and it doesn't impact other projects

[coyotte508]

It should be applied to all read calls, or maybe we can just set it in the MongoClientWithMappings constructor:

Mongoku/src/lib/server/mongo.ts

Lines 132 to 138 in 0292d46

	constructor(url: string, _id: string, name: string) {
	super(url);
	this.url = url;
	this._id = _id;
	this.name = name;
	}

super(url, {readPreference: ...});

(to avoid editing every db call)

[arekborucki]

let me fix it