Tester needed with rclone + crypt + plex to track down bug

I’m trying to fix: https://github.com/ncw/rclone/issues/873 which is most likely a race condition

This seems to go wrong with rclone + crypt + plex so anyone who could test that for me would be most appreciated!

I’ve copied the details here, but if you could attach logs to the issue I’d be grateful.

Thanks

Nick


I made a version of rclone with the race detector enabled (also compiled with go-tip).

http://pub.rclone.org/rclone-v1.34-13-gd8b7156-race.zip

Can you run this with

rclone --log-file race.log  mount ...

Don’t run it with -v or --debug-fuse to keep the logs quiet.

If you give it a thrash with plex to see if you can make it print something in race.log! If you can get it to print a race condition then I’ll be able to fix it most likely!

Note the race detector does slow down rclone a bit.

What kind of output are you looking for? I got a few errors to trigger, but they don’t look anymore descriptive than the original error. These errors were triggers by doing a Library update

My mount command:

root@nas-dev:/tmp/rclone-v1.34-13-gd8b7156-race/linux-amd64# ./rclone --log-file race.log mount acd_crypt: /mnt/cloud/acd-race --allow-other

And the log entries:

2016/11/16 13:14:01 media/video/movie1.avi: ReadFileHandle.Read error: failed to authenticate decrypted block - bad password? 2016/11/16 13:14:54 media/video/movies/movie2.avi: ReadFileHandle.Read error: failed to authenticate decrypted block - bad password?

I’m testing right now, but log’s been quiet, no output yet.

EDIT: Got something.

2016/11/17 05:17:38 movies/foreign/movie.mkv/movie-trailer.mov: ReadFileHandle.Read error: failed to authenticate decrypted block - bad password?

That is the error I’m trying to track down - excellent :slight_smile:

Did you see something like this in the log?

==================
WARNING: DATA RACE
Read at 0x00c4201eee18 by goroutine 13:
  github.com/ncw/rclone/fs.Count.func1()
      /home/ncw/go/src/github.com/ncw/rclone/fs/operations.go:684 +0x56
  github.com/ncw/rclone/fs.ListFn.func1()
      /home/ncw/go/src/github.com/ncw/rclone/fs/operations.go:598 +0x17d

Previous write at 0x00c4201eee18 by goroutine 11:
  github.com/ncw/rclone/fs.Count.func1()
      /home/ncw/go/src/github.com/ncw/rclone/fs/operations.go:684 +0x6f
  github.com/ncw/rclone/fs.ListFn.func1()
      /home/ncw/go/src/github.com/ncw/rclone/fs/operations.go:598 +0x17d

Nothing like that for me so far.

That is a shame - I was hoping the race detector would kick in and point out the problem.

It must be a logic error rather than a race - but where or whose I’m in the dark at the moment!

Just to report in…

I’ve been running Plex with this build doing various Library updates and viewing tests. Unfortunately, I have not seen a race detector message. I’ve been running since 2016/11/16 13:14:01. I have only received various failed to authenticate decrypted block messages.

OK thanks for trying. I’ll try to think of what to try next!

Any update on this issue? And what’s the side effect of it?

I’m looking forward to have acd + crypt + plex be more stable for production system use.

No updates at the moment - I’ve been a bit busy!

I am working on it though - I hate having known bugs in rclone :wink:

Any updates on this issue?

Using the race detector was a bust…

If anyone would like to send me logs of it happening with the latest beta and -v --debug-fuse --dump-headers - email them to nick@craig-wood.com with subject “rclone logs for issue 873” I would be grateful!

I haven’t managed to reproduce the problem is the main stumbling block at the moment.