NodeBB Assets - Object Storage

phenomlab

@razibal created a new bucket and tried again from CLI. It looks fine from here, but everything lands up in /assets/

razibal

aws s3api list-objects-v2 --endpoint-url https://[account].r2.cloudflarestorage.com --bucket nodebb-dev3 | more

{
    "Contents": [
        {
            "Key": "assets/acp.min.js.map",
            "LastModified": "2023-05-17T14:37:46.508Z",
            "ETag": "\"30d96ca6de60c70d482689de4f656784\"",
            "Size": 2115263,
            "StorageClass": "STANDARD"
        },
        {
            "Key": "assets/acp.min.js",
            "LastModified": "2023-05-17T14:37:46.182Z",
            "ETag": "\"ac03fc028d895314c458a64783a5d2df\"",
            "Size": 617833,
            "StorageClass": "STANDARD"
        },

phenomlab

@razibal

{
    "Contents": [
        {
            "Key": "assets/03371bf1d5cbb1eab58e3a0130d1e9c2.js",
            "LastModified": "2023-05-17T17:37:30.453Z",
            "ETag": "\"733f46933105eaf011a37d00a964aa39\"",
            "Size": 502502,
            "StorageClass": "STANDARD"
        },
        {
            "Key": "assets/10822.4553fc0a94f5e4017c9e.min.js",
            "LastModified": "2023-05-17T17:37:29.881Z",
            "ETag": "\"e825b3003d09c6acaef0960484c1cd43\"",
            "Size": 742,
            "StorageClass": "STANDARD"
        },
        {
            "Key": "assets/10943.83723e39cc9dae85e2e7.min.js",
            "LastModified": "2023-05-17T17:37:29.880Z",
            "ETag": "\"a14d728a69056d99f40de4e9d009c14f\"",
            "Size": 1149,
            "StorageClass": "STANDARD"
        },
        {
            "Key": "assets/11307.fd4abf5601dce6fcfaf8.min.js",
            "LastModified": "2023-05-17T17:37:30.211Z",
            "ETag": "\"b81d4a241111819396108c44b1932acc\"",
            "Size": 327,
            "StorageClass": "STANDARD"
        },
        {
            "Key": "assets/11393.036b406d0bb56dc3b1c4.min.js",

A lot more output here, but I think you get the picture.

It looks like everything is being minified into single files?

razibal

just copy a single file as a test

aws s3 cp ./package.json s3://nodebb-dev3/assets/folder1/folder2/package.json --endpoint-url https://[account].r2.cloudflarestorage.com
upload: ./package.json to s3://nodebb-static/assets/folder1/folder2/package.json

aws s3 ls s3://nodebb-dev3/assets/folder1/folder2/package.json --endpoint-url https://[account].r2.cloudflarestorage.co
m
2023-05-17 12:48:58       7920 package.json

phenomlab

@razibal files are being taken from /build/public/ - is that correct ? Shouldn't it just be /public/ ??

razibal

correct

await sync('./build/public', 's3://nodebb-static/assets', {...

phenomlab

@razibal But that is where the minified files are being taken from - and it matches the directory structure.

EDIT - I changed it to just /public/ and it looks much better

razibal

You know the nodebb file layout better than I do, all folders that serve static assets would need to be synched to the bucket

phenomlab

@razibal Yes, I think this is the disconnect. I'm now seeing folders and files.

phenomlab

@phenomlab Hmm - perhaps not. I think you still need ./build/public/ which I am pulling back in now.

EDIT - seems the files are indeed there

However, there is an issue with pulling the files on request.

razibal

so, the script can be modified as

async function syncStaticFiles() {
const folders = ['./build/public', './public']
    await Promise.all(folders.map(async folder => {
    await sync(folder, 's3://nodebb-static/assets', {
        monitor,
        maxConcurrentTransfers: 1000,
        commandInput: {
            ACL: 'public-read',
            ContentType: (syncCommandInput) => mime.lookup(syncCommandInput.Key) || 'text/html'
        }
    });
    }));
    process.exit()
}

razibal

If you sync using the command line, there will be issues as the content-type won't be set. Your last sync for build/pubic was with the command line

phenomlab

@razibal Right.... I think I have it working now with those modifications Running some tests, but all looking good. Will test overnight...

This has a lot of potential using workers and R2, and really needs proper documentation. As you know, it's taken me 4 hours to get this working!

When I run /nodebb build I don't see anything else running afterwards - should it be calling that other script automatically?

phenomlab

@baris @julian I think this should be made into a plugin if I'm honest. Obviously, there are other steps that need to be completed outside of just installing the plugin, but this works very well, and it's extremely fast in my testing within dev.

razibal

@phenomlab said in NodeBB Assets - Object Storage:

This has a lot of potential using workers and R2, and really needs proper documentation. As you know, it's taken me 4 hours to get this working!

Yeah, it only took me a few minutes to get it working as a POC this morning, but , as you experienced, getting that even partially operationalized took a lot longer.

The nice thing about implementing this is that it makes nodeBB essentially infinitely scalable. All static assets are handled by cloudflare workers while the nodeBB API can be deployed to droplets behind a load balancer. Spin up droplets as demand increases and then spin them back down when no longer needed.

phenomlab

@razibal this can probably also be extended to services outside of Cloudflare and AWS.

<baris>

What would go in the nodebb plugin? Seems like all the instructions here are happening outside of nodebb source code, like making a post build script to copy the built assets and the script that runs on cloudflare.

phenomlab

@baris correct, but those lesser conversant could make use of a plugin that completes the NodeBB side.

razibal

@phenomlab said in NodeBB Assets - Object Storage:

@razibal this can probably also be extended to services outside of Cloudflare and AWS.

Thats correct, this would work with any CDN that provides object storage and a mechanism to proxy a path like /assets via cloud workers or CDN proxy service

phenomlab

I thought I'd come back here and post my findings with this so far. My DEV instance is using R2 along with CF workers, and works as expected. Speed is very good.

The production environment uses the traditional method or serving critical assets on site access, but also uses Redis and 3 NodeBB processes (clustering). There is no discernible difference in the speed of the two environments which I find both pleasing, and yet surprising!

I'm performing more tests, and will provide an update here.