"Writing is not allowed after writer was completed", with SignalR 1.1

[viktorgullmark]

Describe the bug

I'm receiving "Writing is not allowed after writer was completed" on the latest release of SignalR with .NET Core 2.2. We previously used 1.0.0-rc1-final without any issues, and all of a sudden started receiving this issue with 1.1 on the client. We also receive a lot of "Task was cancelled" from the RedisHubLifetimeManager. Attaching screenshots below.

We ran SignalR 1.1 on the backend with 1.0.0-rc1-final on the client without any issues, but when we upgraded our client to 1.1, we started receiving this in our logs.

Another change we did while upgrading was to add "skipNegotiate: true" on the client, as well as forcing websockets. Not sure if this is related. I can't reproduce the error locally, since it doesn't happen often and seems related to only a few of our clients.

Also, the server seems to go down quite often after the upgrade. Can't seem to find any reason for this, but it was stable before we upped our clients.

To Reproduce

Can't reproduce the error locally, we just see it in our production environment logs.

Expected behavior

Expected it to be gone with 1.1.0, since I saw previous reports of the problem on RC versions.

Screenshots

Screenshot of writing is not allowed message:

Screenshot of task cancelled exception:

Additional context

dotnet --info output:

This is on our production environment. Locally we have the 2.2 runtime.

EDIT: If this issue is misplaced or inproper, feel free to close it. I didn't know where to turn.

[BrennanConroy]

The task canceled exception from Redis occurs when you try to add or remove a connection from a group where the connection either doesn't exist, or is on another server and the other server took too long to respond.

As for the "Writing not allowed" exception, how often are you seeing this? Do the server logs show that a client was disconnected around the same time the exception occurs?

[viktorgullmark]

Thanks for the input. We've resolved the Task canceled exception now thanks to your clarification.

We're seeing the "Writing is not allowed after writer was completed" a few times every hour, with around 100 users online. Can't find a disconnect at this timestamp in the logs though 🤔

[BrennanConroy]

Could you capture logs around the exception, at Debug level preferably.

[bradygaster]

Closing this as we haven't heard from you. Please feel free to comment if you're able to get the information we're looking for and we can reopen the issue if we're able to identify the problem.

[imperugo]

Hi,
I'm getting the same error with this scenario:

.NET Core 2.2, SignalR 1.1, scaling using Redis (2.x) and hosting into Docker.

Unfortunately, I'm unable to reproduce the problem and the log isn't so helpful (Writing is not allowed after writer was completed)
Is there anything I can do to provide more information? (add more logging somewhere, try another build or whatever?).

I'm also using Groups so, probably is the same problem of @viktorgullmark but I don't have the issue related to the Task Canceled exception.

Let me know.

[codingsteff]

We have the same errors, only in production inside Docker behind Nginx.
The messages appeared after upgrade to .NET Core 2.2.

It doesn't matter if we use SignalR 1.1 or 1.0 on the Client Side.

By the way, we don't use Groups or Redis.

[BrennanConroy]

@imperugo Enabling Debug level logs would be nice. I'd like to see the logs around the failure if possible, as they can give context on what is happening on the server.

@codingsteff Can you confirm that switching from SignalR 1.0.0 to 1.1.0 on the server causes the issue? And could you also collect logs please?

[imperugo]

@BrennanConroy I'll switch the log only for "Microsoft.AspNetCore.SignalR.*" to debug.

Soon news