debian-mirror-gitlab/doc/administration/static_objects_external_storage.md

stage	group	info	type
Create	Editor	To determine the technical writer assigned to the Stage/Group associated with this page, see https://about.gitlab.com/handbook/engineering/ux/technical-writing/#assignments	reference

Static objects external storage

Introduced in GitLab 12.3.

GitLab can be configured to serve repository static objects (for example, archives or raw blobs) from an external storage, such as a Content Delivery Network (CDN).

Configuring

To configure external storage for static objects:

1. Navigate to Admin Area > Settings > Repository.
2. Expand the Repository static objects section.
3. Enter the base URL and an arbitrary token. When you set up external storage, you'll use a script that uses these values as ORIGIN_HOSTNAME and STORAGE_TOKEN.

The token is required to distinguish requests coming from the external storage, so users don't circumvent the external storage and go for the application directly. The token is expected to be set in the X-Gitlab-External-Storage-Token header in requests originating from the external storage.

Serving private static objects

GitLab will append a user-specific token for static object URLs that belong to private projects, so an external storage can be authenticated on behalf of the user. When processing requests originating from the external storage, GitLab will look for the token in the token query parameter or in the X-Gitlab-Static-Object-Token header to check the user's ability to access the requested object.

Requests flow example

The following example shows a sequence of requests and responses between the user, GitLab, and the CDN:

sequenceDiagram
    User->>GitLab: GET /project/-/archive/master.zip
    GitLab->>User: 302 Found
    Note over User,GitLab: Location: https://cdn.com/project/-/archive/master.zip?token=secure-user-token
    User->>CDN: GET /project/-/archive/master.zip?token=secure-user-token
    alt object not in cache
      CDN->>GitLab: GET /project/-/archive/master.zip
      Note over CDN,GitLab: X-Gitlab-External-Storage-Token: secure-cdn-token<br/>X-Gitlab-Static-Object-Token: secure-user-token
      GitLab->>CDN: 200 OK
      CDN->>User: master.zip
    else object in cache
      CDN->>GitLab: GET /project/-/archive/master.zip
      Note over CDN,GitLab: X-Gitlab-External-Storage-Token: secure-cdn-token<br/>X-Gitlab-Static-Object-Token: secure-user-token<br/>If-None-Match: etag-value
      GitLab->>CDN: 304 Not Modified
      CDN->>User: master.zip
    end

Set up external storage

While this procedure uses Cloudflare Workers for external storage, other CDNs or Function as a Service (FaaS) systems should work using the same principles.

1. Choose a Cloudflare Worker domain if you haven't done so already.

2. In the following script, set the following values for the first two constants:

   • ORIGIN_HOSTNAME: the hostname of your GitLab installation.

   • STORAGE_TOKEN: any arbitrary secure token (e.g. you can get one by running pwgen -cn1 64 on a UNIX machine). Save this token for the admin panel, as described in the configuring section.

     const ORIGIN_HOSTNAME = 'gitlab.installation.com' // FIXME: SET CORRECT VALUE
     const STORAGE_TOKEN = 'very-secure-token' // FIXME: SET CORRECT VALUE
     const CACHE_PRIVATE_OBJECTS = false
     
     const CORS_HEADERS = {
       'Access-Control-Allow-Origin': '*',
       'Access-Control-Allow-Methods': 'GET, HEAD, OPTIONS',
       'Access-Control-Allow-Headers': 'X-Csrf-Token, X-Requested-With',
     }
     
     self.addEventListener('fetch', event => event.respondWith(handle(event)))
     
     async function handle(event) {
       try {
         let response = await verifyAndHandle(event);
     
         // responses returned from cache are immutable, so we recreate them
         // to set CORS headers
         response = new Response(response.body, response)
         response.headers.set('Access-Control-Allow-Origin', '*')
     
         return response
       } catch (e) {
         return new Response('An error occurred!', {status: e.statusCode || 500})
       }
     }
     
     async function verifyAndHandle(event) {
       if (!validRequest(event.request)) {
         return new Response(null, {status: 400})
       }
     
       if (event.request.method === 'OPTIONS') {
         return handleOptions(event.request)
       }
     
       return handleRequest(event)
     }
     
     function handleOptions(request) {
       // Make sure the necessary headers are present
       // for this to be a valid pre-flight request
       if (
         request.headers.get('Origin') !== null &&
         request.headers.get('Access-Control-Request-Method') !== null &&
         request.headers.get('Access-Control-Request-Headers') !== null
       ) {
         // Handle CORS pre-flight request
         return new Response(null, {
           headers: CORS_HEADERS,
         })
       } else {
         // Handle standard OPTIONS request
         return new Response(null, {
           headers: {
             Allow: 'GET, HEAD, OPTIONS',
           },
         })
       }
     }
     
     async function handleRequest(event) {
       let cache = caches.default
       let url = new URL(event.request.url)
       let static_object_token = url.searchParams.get('token')
       let headers = new Headers(event.request.headers)
     
       url.host = ORIGIN_HOSTNAME
       url = normalizeQuery(url)
     
       headers.set('X-Gitlab-External-Storage-Token', STORAGE_TOKEN)
       if (static_object_token !== null) {
         headers.set('X-Gitlab-Static-Object-Token', static_object_token)
       }
     
       let request = new Request(url, { headers: headers })
       let cached_response = await cache.match(request)
       let is_conditional_header_set = headers.has('If-None-Match')
     
       if (cached_response) {
         return cached_response
       }
     
       // We don't want to override If-None-Match that is set on the original request
       if (cached_response && !is_conditional_header_set) {
         headers.set('If-None-Match', cached_response.headers.get('ETag'))
       }
     
       let response = await fetch(request, {
         headers: headers,
         redirect: 'manual'
       })
     
       if (response.status == 304) {
         if (is_conditional_header_set) {
           return response
         } else {
           return cached_response
         }
       } else if (response.ok) {
         response = new Response(response.body, response)
     
         // cache.put will never cache any response with a Set-Cookie header
         response.headers.delete('Set-Cookie')
     
         if (CACHE_PRIVATE_OBJECTS) {
           response.headers.delete('Cache-Control')
         }
     
         event.waitUntil(cache.put(request, response.clone()))
       }
     
       return response
     }
     
     function normalizeQuery(url) {
       let searchParams = url.searchParams
       url = new URL(url.toString().split('?')[0])
     
       if (url.pathname.includes('/raw/')) {
         let inline = searchParams.get('inline')
     
         if (inline == 'false' || inline == 'true') {
           url.searchParams.set('inline', inline)
         }
       } else if (url.pathname.includes('/-/archive/')) {
         let append_sha = searchParams.get('append_sha')
         let path = searchParams.get('path')
     
         if (append_sha == 'false' || append_sha == 'true') {
           url.searchParams.set('append_sha', append_sha)
         }
         if (path) {
           url.searchParams.set('path', path)
         }
       }
     
       return url
     }
     
     function validRequest(request) {
       let url = new URL(request.url)
       let path = url.pathname
     
       if (/^(.+)(\/raw\/|\/-\/archive\/)/.test(path)) {
         return true
       }
     
       return false
     }
     

3. Create a new worker with this script.

4. Copy your values for ORIGIN_HOSTNAME and STORAGE_TOKEN. Use those values to configure external storage for static objects.