[mythtv-users] trunk issue

[David Shay]

On Thu, Aug 19, 2010 at 1:01 AM, Michael T. Dean <mtdean at thirdcontact.com>wrote:

>  On 08/18/2010 11:34 PM, David Shay wrote:
>
>> Running trunk 25701 when I got the errors below. Backend was
>> psuedo-crashed/deadlocked or at least unresponsive to my frontend. I have
>> 3
>> backends and 1 frontend only. The stream below included some commflagging,
>> not sure if that was a part of the problem or not.
>>
>>
>> MythSocket(1e659e0:11): readStringList: Error, timed out after 7000 ms.
>> MythCoreContext, Error: Timeout connecting server socket to master backend
>> Connection to master server timed out.
>> Either the server is down or the master server settings
>> in mythtv-settings does not contain the proper IP address
>>
>> CommFlagFast: 0
>> special_decode: 0
>> AFD Warning: ScanATSCCaptionStreams() called with no PMT
>> AFD: Opened codec 0x7faff40132f0, id(MPEG2VIDEO) type(Video)
>> AFD: codec MP2 has 2 channels
>> AFD: Opened codec 0x7faff4004d50, id(MP2) type(Audio)
>> MythCoreContext: Connecting to backend server: 192.168.1.93:6543 (try 1
>> of
>> 1)
>> AFD Warning: ScanATSCCaptionStreams() called with no PMT
>> [mpeg2video @ 0x7fb00c93c4a0]warning: first frame is no keyframe
>> [mpeg2video @ 0x7fb00c93c4a0]warning: first frame is no keyframe
>> MythSocket(1f60b80:10): readStringList: Error, timed out after 7000 ms.
>> Protocol version check failure.
>> The response to MYTH_PROTO_VERSION was empty.
>> This happens when the backend is too busy to respond,
>> or has deadlocked in due to bugs or hardware failure.
>> Preview: Grabbed preview '/var/data/tv/data3/1222_20100815000000.mpg'
>> 720x480 at 3600s
>> MythCoreContext: Connecting to backend server: 192.168.1.93:6543 (try 1
>> of
>> 1)
>> MythSocket(1e62b60:10): readStringList: Error, timed out after 7000 ms.
>> Protocol version check failure.
>> The response to MYTH_PROTO_VERSION was empty.
>> This happens when the backend is too busy to respond,
>>
>
> I've seen this happen in 0.23-fixes when the MythXML/http server is
> "overstressed"--i.e. I can cause it on demand by deleting all the PNG
> thumbnails from my backend servers and then going to the MythWeb Recorded
> Programs page (and it will fail after some huge number of previews are
> generated).  Unfortunately, I haven't looked into tracking it down since
> I've been busy with other stuff and (at least on my system) it's easy enough
> to avoid triggering it by not doing that.
>
>
This was @ 3 AM with no end user interaction at all (no one on Mythweb, no
other frontends running), so I'm not sure what else might overstress the XML
server. Was previously running release .23 but had some audio sync issues
when commercial skip activated on HDHR recorded content. Those are fixed I
think, but this is quite bad if it kills the backend. Could cross-backend
commercial-flagging potentially cause the overloading? Would still have to
be a new regression since release .23.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mythtv.org/pipermail/mythtv-users/attachments/20100819/a18a781c/attachment.htm>