[mythtv-users] Recording issues with I/O load

warpme warpme at o2.pl
Wed Sep 21 13:47:15 UTC 2011


Hi,
I'm not sure did I get all correctly, but...

>Well, when the new system runs for itself, theres no problem at all,
>recordings are pretty well. But as soon as I'm starting streaming an
>archived recording from the RAID to my TV (no matter if I do this via
>NFS, UPnP or Samba) the recordings get bad. There are a lot of artefacts
>and audio glitches in the recording.

I understood starting read operations on RAID + network operations on NIC causes
distortions on DVB recordings, right ?

>As soon as I stop streaming this
>'noise' vanishes immediately. The same happens, when I'm just copying
>an archived recording to my notebook for example, or when I'm md5summing
>any archived video-file on the RAID.

This tells me READ operation on RAID are triggers here.
If You refer to md5summing - is it on local console or on LAN ?

> It happens every time when I
>generate any mentionable I/O load to the system.
>It happens not, when I'm copying Files from external USB-HDD to the RAID.

This means that issue is related exclusively to RAID read operations, but not RAID write operations, right ? 

>I've searched this list and other for an explanation to this, but what
>I've found on this topic are really old threads like this one:
>http://www.gossamer-threads.com/lists/mythtv/users/159591 or this one:
>http://pchdtv.com/forum/viewtopic.php?p=6378
>These guy's troubles seem to be at least partially to be the same as
>mine, except for the fact, that I have different hardware. In these and
>other threads people are talking about RF-noise or EMI causing bad
>reception to the tuner of the DVB card, when there's a higher I/O load
>to the system. Others are suggesting playing with the PCI-latency as
>described here: http://www.mythtv.org/wiki/PCI_Latency

>What I did:
>- I've tried playing around with the latency but without any effect. As
>.far as I'm understanding this, the SATA-controller is connected to the
>PCIe, which is independent from the PCI-bus which holds the DVB-adaptor.
>So it makes really sense to me that playing with the latency doesn't
>have any effect on this issue.
>- I've observed the reception level using femon while recording and
>.streaming the same time -> didn't see anything suspicious.
>- I deactivated the onboard Realtek R8168 in BIOS and replaced it with
>an Intel PCIe GBit card -> no effekt
>- I shut down mythtv-backend and tried DVB live-tv with me-tv -> same
>effekt, picture gets corrupted

Last test means that issue is not related to mythtv.

Lack of issue when you are in RAID read operations means it is not EMC as I can't imagine so specific difference in EMC radiation between read & write operations that it will cause so binary effect (write:no any artifacts, read: many artifacts).

Currently as root cause I would consider 2 areas:

A\RAID read & write interrupts service / RAID DMA support implementation.
I can imagine specific implementation where Your RAID module has event driven interrupts for write operations and pooling based for read operations (by bug or by bad design). I can imagine effect of blocked/missed transfer requests from DVB card by ongoing read operation on RAID.
Another theory is that read operation is wrongly programming DMA subsystem for read transactions, so parallel DVB DMA transactions can't be properly or timely realized.  
For testing this hypothesis - may You temporarily rearrange Your storage subsystem that DVB recordings are written & read to/from the same storage, but without components responsible for RAID ?
I would suggest to do test where You keep ALL HW the same, and only rearrange software that RAID is not used. 
If it will be OK - then issue is in Your RAID implementation. Then You can look more closely in i.e. interrupts. Are they shared. Do Your system uses MSI ?

If it will be still bad - from many possible, as first I would consider 2 next areas:

B\Start to make sure issue isn't relate to i.e buggy HDD firmware (e.g. bad bus-mastering transactions for read operations) or issue with SATA path (cable, controller, etc). Here You can i.e. temporarily replace HDD a see how it goes. If NOK, look more closely on SATA.

C\Power related issues. I can imagine read operation on HDD which by interference can trigger Your power supply into resonance (it might be when Your power supply pulse converter has feedback loop prone to rising resonance if load has v.specific ripple nature). Sounds like absurd, but many today's electronic has quickly drying capacitors and this can cause v.strange issues. I saw such issues in real live, believe me.

Good luck         


More information about the mythtv-users mailing list