Latest master: first time wizard doesn't proceed

[lbeltrame]

Describe the bug
I updated to the latest master (to follow what's going on, I'm prepared to live with the consequences) and I started fresh with a new data dir.
However the setup locked up right after selecting the language in the first time wizard with a few errors in the logs (see below).

To Reproduce

1. Build latest master state
2. Start with a clean data directory

Expected behavior
First time wizard proceeds.

Logs

[2019-03-09 15:45:50.395 +01:00] [INF] ExecuteQueuedTasks
[2019-03-09 15:45:53.008 +01:00] [ERR] Access token is invalid or expired.
[2019-03-09 15:45:53.012 +01:00] [ERR] Error processing request
System.ArgumentNullException: Value cannot be null.
Parameter name: user
   at MediaBrowser.Controller.Entities.Folder.GetChildren(User user, Boolean includeLinkedChildren, InternalItemsQuery query) in /home/jellyfin/jellyfin/MediaBrowser.Cont
roller/Entities/Folder.cs:line 1161
   at Emby.Server.Implementations.Library.UserViewManager.GetUserViews(UserViewQuery query) in /home/jellyfin/jellyfin/Emby.Server.Implementations/Library/UserViewManager
.cs:line 44
   at MediaBrowser.Api.UserLibrary.UserViewsService.Get(GetUserViews request) in /home/jellyfin/jellyfin/MediaBrowser.Api/UserLibrary/UserViewsService.cs:line 84
   at Emby.Server.Implementations.Services.ServiceExecGeneral.Execute(Type serviceType, IRequest request, Object instance, Object requestDto, String requestName) in /home
/jellyfin/jellyfin/Emby.Server.Implementations/Services/ServiceExec.cs:line 88
   at Emby.Server.Implementations.Services.ServiceController.Execute(HttpListenerHost appHost, Object requestDto, IRequest req) in /home/jellyfin/jellyfin/Emby.Server.Imp
lementations/Services/ServiceController.cs:line 177
   at Emby.Server.Implementations.Services.ServiceHandler.ProcessRequestAsync(HttpListenerHost appHost, IRequest httpReq, IResponse httpRes, ILogger logger, String operationName, CancellationToken cancellationToken) in /home/jellyfin/jellyfin/Emby.Server.Implementations/Services/ServiceHandler.cs:line 95
   at Emby.Server.Implementations.HttpServer.HttpListenerHost.RequestHandler(IHttpRequest httpReq, String urlString, String host, String localPath, CancellationToken cancellationToken) in /home/jellyfin/jellyfin/Emby.Server.Implementations/HttpServer/HttpListenerHost.cs:line 575
[2019-03-09 15:45:53.038 +01:00] [ERR] Access token is invalid or expired.

Also the logs are full of [WRN] Unable to determine session based on url like this: http://example.com/embywebsocket?apikey=myapikey`

System (please complete the following information):

• OS: openSUSE Tumbleweed
• Jellyfin reverse proxied through nginx
• Jellyfin Version: master

[anthonylavado]

Yeah. @cvium is working on this in #1079.

[lbeltrame]

I asume there's no way to bypass the first time wizard, either documented or not?

[lbeltrame]

I was able to proceed further without using the reverse proxy (more errors on Connection Upgrade) when manually applying #1079. However, with a reverse proxy it is impossible to log in now as the server returns 400 (Bad Request). The error is as follows:

mar 17 22:22:38 sasara.home dotnet[9170]: [22:22:38] [INF] Connection id "0HLLB3HMRVHT5" bad request data: "Requests with 'Connection: Upgrade' cannot have content in the request body."

[lbeltrame]

Hm, it looks like a case of aspnet/KestrelHttpServer#1263

Adding the header suggested there fixes it. Which means that for reverse proxying Jellyfin with this change (#1079) you need:

        proxy_pass <whatever is needed>
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-for $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Protocol $scheme;
        proxy_set_header X-Forwarded-Proto $scheme;
        proxy_set_header X-Forwarded-Host $http_host;
        proxy_redirect off;

        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection $http_connection;

In your reverse proxy configuration in case of nginx.

These changes + #1079 fix the current issue.

[JustAMan]

@lbeltrame what new Nginx headers did you apply? Related to websocket or X-* stuff?

[lbeltrame]

I added:

• proxy_set_header X-Forwarded-Host $http_host;
• proxy_set_header Connection $http_connection (as opposed to "upgrade")
• proxy_set_header X-Real-IP $remote_addr

[JustAMan]

@lbeltrame can you check that first time wizard proceeds normally without any of configuration changes if you access Jellyfin directly (i.e. not via reverse proxy)?

[lbeltrame]

Ok, it looks like those errors in the log as previously reported were deeply misleading and in fact it was only a matter of reverse proxying things differently. I started with a fresh configuration and with those extra bits the wizard proceeded (it did proceed also without proxying, but that's a no-go for me).

This means that they ought to be added to the docs (will think about a PR if no one beats me to it).

The "null" exception still occurs BTW (however the error about the access token is no longer there), but it doesn't look critical.

FTR, even if #1079 wasn't the solution for the issue, it was very helpful (due to its errors) in actually finding out the cause of the problem.

[JustAMan]

@lbeltrame I've created an issue on jf-docs repository, feel free to solve it via PR if you have time :)

[joshuaboniface]

Should be fixed in 10.3.0 release.