TSreader appears to have a lot of correction to prediction errors (1 Viewer)

Benoire

MP Donator
  • Premium Supporter
  • March 17, 2012
    679
    161
    44
    Auckland
    Home Country
    New Zealand New Zealand
    @Owlsroost

    Sorry to pull you in, but you appear to be the TSReader.ax guru.

    Bit of background first. I've got a Server 2012 R2 setup with BDA installed and this runs TVServer without a hitch, it can read and write to the HDD including timeshifting via UNC or RTSP. I know this setup isn't supported so that's cool but I'm having difficulty with TSreader on the main HTPC and it losing what appears to be sync with the timeshift file on the server. I've attached the logs from the HTPC for reference. I'm also running @DJBlu modifications however in this instance, TSreader isn't amended from what I can tell so it is straight up MP client (with a quick record option, so a slight mod to the code).

    Effectively UNC is the way I want to connect as RTSP has around an 8s connection time, where as UNC is only around 4 from pressing 'go' on the remote and the channel playing. The issue comes sometime in to the playback when the picture starts to freeze and pause before stopping entirely (with a static frame). looking at the logs it states that correction to prediction has occurred and that's when it gets stuck.

    I'm running your 78v2 tsreader.ax from another thread as a trial but realistically it seems to occur no matter which tsreader I use.

    Is there anything obvious in the logs that would show why it might be doing that?

    Any help you provide will be great!

    Chris
     
    Last edited:

    Owlsroost

    Retired Team Member
  • Premium Supporter
  • October 28, 2008
    5,540
    5,038
    Cambridge
    Home Country
    United Kingdom United Kingdom
    These lines in the log:

    [2013-11-13 21:09:13,692] [1ac52008] [ 540] - FileReader::Read() read failed - error = -2146893818
    [2013-11-13 21:09:13,693] [1ac52008] [ 540] - MultiFileReader has error 0x80 in Loop 0. Try to clear SMB Cache.

    ...mean it can't read the remote timeshift buffer files (the 'correction to predicted duration' messages are because the file duration has stopped increasing - as expected in this situation). The error code seems to be some sort of security error - NTE_BAD_SIGNATURE - try putting 'HRESULT 0x80090006' into google (-2146893818 = 0x80090006 in hex) and see if it gives you any clues (I'm not familiar with windows server etc) - http://blogs.msdn.com/b/eldar/archive/2007/04/03/a-lot-of-hresult-codes.aspx
     

    Benoire

    MP Donator
  • Premium Supporter
  • March 17, 2012
    679
    161
    44
    Auckland
    Home Country
    New Zealand New Zealand
    @Owlsroost

    So from what I can gather, the error codes are due to the windows 7 HTPC losing access to the SMB share from the Server. How long would it take for MP to freeze if it can't find the file growing? Could the converse had happened and that the TVserver had stopped writing the timeshift buffer for some reason, altough the logs from the TVserver simply show it continuining on as normal with no stop event until the client issues the stop event.

    I've got two other options to play with later regarding the timeout length for UNC connections and also mapping as a 'local drive'. Either way this is interesting as Server 2012 Essentials doesn't appear to suffer from this but Server 2012 (R1 or R2) does, and that includes R2 running the esentials role.

    I'm going to test Argus Recorder as I know that has a lot more buffering, sync logic than TSwriter and see if it is the TVEngine causing the trouble...

    I don't know why, when setup similar to Essentials with the BDA drivers from Windows 8.1, would it act differently to essentials.

    Again I'm just sounding off you as I know my particular setup is not supported and nor do I ask it to be, but I'm interested in helping improve the MP experience by running out of spec stuff (virtualised OS's on modern hardware etc etc.).

    @DJBlu you didn't make any other modifications to TSwriter apart from to support the custom data grabber? Nothing that should affect the way the timeshift files should grow?

    The more I type, the more I think it it a network connectivity issue when running connected to a domain...
     

    Benoire

    MP Donator
  • Premium Supporter
  • March 17, 2012
    679
    161
    44
    Auckland
    Home Country
    New Zealand New Zealand
    That's cool :) If I get this working fine I'll try and add it to the Wiki for other peoples reference if they need it... It's always good to have this sort of knowledge recorded somewhere...

    Thanks for your thoughts so far :)
     

    Benoire

    MP Donator
  • Premium Supporter
  • March 17, 2012
    679
    161
    44
    Auckland
    Home Country
    New Zealand New Zealand
    Right... So I might have nailed it. Still testing so far but got 3 tuners active (2 timeshift separate clients and 1 recording).

    I was starting to believe it was related to the virtual network card/switch that is used when virtualizing software so I decided to reinstall Server 2012 R2 using the eval version (to save using my key and causing more issues down the track). Installing the Windows 8.1 BDA files allowed full use of the tuners. As expected TVserver was fine, no issues... However as Windows is now natively accessing the network card it seems that it is stable. This makes me believe that while UNC is fine over virtualized switches, it throws out the timing of video which requires fairly precise stuff.

    My next test is to enable Hyper-v, create a virtual switch and then install a new server. IF I start to get the issues as before then I know is virtual switches/Ethernet adapters which MP over UNC cannot deal with.

    I'm lucky that my motherboard/cpu allows me to pass through any attached PCi/PCi-e device to a virtual machine so I do expect to solve this by purchasing a new Ethernet card and then pass that through to the VM.

    I'll continue to update this thread as I explore further and will update (or provide an update) for the wiki so other people can understand the benefits/pitfalls of virtualizing their hardware when running TVserver (and this might apply to the centralized server approach of MP2 as well).

    Chris
     

    Benoire

    MP Donator
  • Premium Supporter
  • March 17, 2012
    679
    161
    44
    Auckland
    Home Country
    New Zealand New Zealand
    I would agree with that however i do not have the space for more than one server tower at present. Eventually i will move to rack mounts but virtualised has its benefits for me right now.
     

    Benoire

    MP Donator
  • Premium Supporter
  • March 17, 2012
    679
    161
    44
    Auckland
    Home Country
    New Zealand New Zealand
    So it lives! Server 2012 R2 as base with bda and tuner drivers and then hyper-v with a vswitch added. Difference here is that the base os still has direct access to the Ethernet device where as in esxi only the host has access and everything else is virtualised.

    I think to make it work in esxi would require a pcie Ethernet card passed directly to the vm as long as the motherboard supports Iommu or similar.
     

    Users who are viewing this thread

    Top Bottom