@nennogabriel said in Directus fail to create an user from Cloudron Open ID:

the new version of Directus on cloudron dos not use LDAP and you cannot use OPENID but CLOUDRON as variable.

what do you mean by this? Why can we not use OPENID?

I could successfully connect to a test postgres database with these environment variables:

DB_HOST=IP
DB_PORT=Port
DB_DATABASE="db_name"
DB_USER="user_name"
DB_PASSWORD="password"

Then I ran
npx directus bootstrap
to initialise the database.

When I run commands like
npx directus count table_name
it provides correct answers indicating that the database connection works.

However, for some reason, the database doesn't show up in my GUI.

I was able to deploy with cloudron my nextjs frontend app fetching my directus backend via API here is the repository with STEP-BY-STEP guide: https://github.com/alexmazaltov/nextjs-app-for-directus-vai-cloudron

@girish said in Issue with very high memory usage:

so it's not related to cloudron end.

not sure about this.

When i create a Directus project on a VPS using docker (docker compose or via Portainer), with almost identical configuration as the Cloudron Directus project, i don't get this issue.

Example: Container memory usage when uploading a 8GB file to Directus, barely goes up by some 200MB, whereas in Cloudron, the memory always gets consumed by the entire size of the file (8GB in this case).

It is only created within the directus app instance. Different apps run isolated from one another on Cloudron. So if I understand you correctly, you are trying to build your own custom app next to directus? If so you need to set the source your own /app/data/env.sh with the export of the env var.

This should be fixed now with latest package. Unfortunately I found that if a previous extension installation failed, the database would contain that info, making it impossible to install the same extension again. So you may have to manually delete records from the database to get over this 😕

There are some further notes in https://docs.directus.io/getting-started/architecture.html#versioning . For them, "Minor — A new feature or meaningful improvement to an existing feature: may or may not be breaking."

I think Cloudron packaging has to follow what upstream does, otherwise, every minor release now becomes a major in Cloudron. Maybe best to turn off automatic updates for Directus and analyze each Cloudron package update?

@shrey said in App status stuck on "Starting...":

ECONNREFUSED

This is normal. When the app is still starting up, it's not responding to health check requests and refusing connection. As long as it becomes 'Running' eventually, all is good.

Seems to work like a charm!

@micmc I worked through this and got everything working in the past. I had to install the directus-cli on my local machine and use something like Visual Studio Code or similar program to connect to my Directus instance hosted inside Cloudron, to run the AgencyOS importer.

Separate from that, I forked the "AgencyOS" code base to a private Github repo and published using Vercel for the frontend.

Based on my experience, because anything outside of /app/data/ on Cloudron is read only, I had to break apart the AgencyOS frontend and then just connect back to Directus.

I'd be aware, AgencyOS is still under active development and it was created as more of a proof of concept, so in my experience you're bound have quite a few issues. But I got it working in the past, but take lots and lots of backups for both the frontend and your directus instance.

@samsls Thanks for following up with me on this!

I used the Directus CLI with a Token for the admin account to import the data into a "default vanilla installation of Directus". Everything imported without any issues. Afterwards, I deployed the front end using Vercel from a private GitHub repo.

Based on the reply from @girish, I put some additional configuration into the env.sh file and restarted Directus.

I've placed that into the env.sh as a wide-open test to at least see if I could get it working, but no luck.

I know there are some other options, that you can force some of these settings into your Vercel deployment using the vercel.json file, but haven't gotten to that yet.

There are some open issues that it seems they're looking into based on their GitHub.

Yes seems to work thank you!

Thank you for helping me make it work.

Strange this .mjs thing with them, I'm going to open an issue with them.

Those are only container upper limits, not what the apps will use.

You can significantly increase the size of the swap file for more headroom, but things will slow down while swapping. You'll want more RAM as certain apps will use a lot more when you start heavier use.

Cost wise, there are server providers that max RAM such as Contabo and SSDNodes at a decent price.

Directus is using postgres on Cloudron, see https://git.cloudron.io/cloudron/directus-app/-/blob/master/CloudronManifest.json#L15

Regarding the ping, took me a bit to realize that you have to hover over that wifi signal strength icon. However on a fresh install here on a not so beefy server it shows values around 50ms. But more importantly I am not actually sure what this represents and what the usefulness of this is.

I've found this:
https://docs.cloudron.io/apps/directus/

Example Link is not working:
https://github.com/directus/directus/blob/main/api/example.env

@pbischoff said in Scaling/vCPU core usage of Directus in Cloudron:

But this is a problem for almost every node.js app of the cloudron App Store, isn't it? Do they all use only one vCPU core?

Apps like NodeBB (this forum) support so called cluster mode - https://nodejs.org/api/cluster.html . Essentially, this creates worker threads to spread load across CPU. This is part of nodejs.

But to keep in mind, nodejs by itself can handle quite a bit of parallelism with just one CPU. Most of the work done by the main process is just http request processing. This is usually I/O bound and not CPU bound. Nodejs being asynchronous can handle requests in parallel in just one process quite easily. We are talking about 1000s of requests a second.

Many other nodejs apps like Cloudron code itself, peertube, offload the processing of static assets to a side installation of nginx. So, when downloading images/css files etc, it doesn't even hit nodejs. When node code has to do heavy work, it usually just puts it in a queue and the queue handler is a separate process (which can thus be restarted and killed easily). The queue handler will use another cpu core.

For directus, I am not an expert on the app. But, you want to maybe use some worker thread or queue or something like that to make use of more CPUs.

@girish said in Unable to connect to a DigitalOcean managed database:

None of the feature like backup/restore will work

So far, they've been working fine. Because, all that's required to be backed up in such cases is the app code. Of course, as the database is external, it is accepted that the backup system is external & independent to Cloudron.

@girish said in Unable to connect to a DigitalOcean managed database:

Not an expert on the apps and I don't know if it's the main purpose. But they work just fine with fresh databases as well where you use it like a nocode/locode app.

Yes, pairing up with external databases is indeed, one of the primary features of these apps. And using a fresh database is just one of the features, and not always preferred.

You might want to state clearly and prominently, particularly in the cases of such apps, that external connections are not allowed/recommended.