[mythtv-users] Tips for user jobs on NFS mounts over Wi-Fi?

[Michael Watson]

On 6/05/2015 9:33 AM, Dave MythTV wrote:
> Hello everyone.
>
> Of course a fast wired network would be greatly preferred, but does 
> anyone have any tips for running high-bandwidth user jobs 
> (transcoding, etc) on remote backends which are connected via a Wi-Fi 
> network?
>
>
>
> My primary MythTV machine is a Frontend/Backend combo, with all of the 
> tuners, storage, etc, and it is working great.  However, I'm working 
> on a side project to transcode many years of recordings (~3 TB) to 
> H.264 for longer term archival and better compatibility with our 
> tablets and other portable devices.
>
> In doing this, I'm going to be temporarily running tunerless remote 
> backends (really just mythjobqueue) on several other machines in the 
> house to help work through the backlog.  These machines are much more 
> powerful (2-5x so), but will only have a Wi-Fi connection to the 
> master backend with the storage drives.
>
>
> I have one remote backend up and running, but I'm starting to suspect 
> that network I/O will ultimately limit the scalability and overall 
> transcoding speed - particularly once I get a second remote backend 
> configured.
>
>
> I'm thinking of modifying my user job scripts, so that instead of 
> transcoding directly to/from the NFS mounted drive, they instead copy 
> the entire recording to local storage, do the transcode locally, and 
> then copy the entire transcoded result back to the NFS share.   My 
> theory is that this will make the network transfer more efficient, and 
> prevent the transcoding processes from stalling constantly while 
> waiting on network I/O... but it's just a theory.  I could also see it 
> being slower in some cases, as it has to wait for the entire file to 
> be received before it can start any of the CPU-heavy transcoding 
> work.  So I'm not sure which would be more effective.
>
>
> Has anyone worked within this situation, and do you have any advice?  :-)
>
> Thanks
> - Dave
>
Yes, more than one transcode job is likely to saturate the wifi network, 
causing issues.
One way I can think of is;

Master Backend Script:
     Copy File to Be Transocde to "/Convert" Directory.
     Wait for Transcode to be completed.  (Wait for 
"transcodefilename.complete" or "transcodefilename.error")
     Copy Transcoded file back to storage
     Remove old files, and status files
     Update DB with new filename, clear/update seek tables etc.

Slave Backend;  (Do not need to be Myth Slave Backends)
      Poll "/Convert" Directory for jobs
      Create file in "/Convert" Directory to claim ownership of job
      Copy File from Master
      Transcode File
      Copy transcoded file back to Master
      Create "transcodefilename.complete" or "transcodefilename.error" 
to tell master backend of status.