SlimServer 2004_09_23-1 and AlienBBC 0.9 problems

[Triode]

Hi Jules,

> Can you try a test for me? Restart the server, proxy, and kill off any
> alienstream zombies. Start the proxy in a separate shell, so that you can
> see it's output (when you first start it, it has nothing to say... just
> leave it there). Then start slimserver, and play a stream. Try stopping
> the
> stream in a bit, and see if the proxy notices and tries to clean up. If it
> does, check that alienstream isn't still running anyway. If that works,
> then
> we'll need to work out why you're getting zombies piling up.

Please see attached for logs - outputs from top and from the proxy.

I should say from the start that my server is a 200 MHz Pentium with 96M
memory and Mandrake 10. It runs slimserver fine with PCM streaming from
FLAC etc but stalls with concurrent access to the web interface [I turn off
web refresh]. So I may be stretching this more than others and it may just
be lack of CPU.

My observations:
1) CPU load of the transcoder is independant of how many alienstream
processes are created. Not that high CPU load is seen before any streams
are started and is highest when nothing else is going on. I wonder if this
is just the interaction of the timer you mentioned and my CPU speed. Any
suggestion of whether this can be increased? CPU load of this process also
seems to consume any spare cycles upto 70 odd percent..

2) Although top shows multiple alienstreams, this may be normal? They seem
to start and stop together and most of the CPU time goes against one of
them?

Will look at mplayer now...

Thanks again,

Adrian

[Jules Taplin]

Hi Triode.

Just stopped by my machine for 2 mins. I'll try to look at this better
later.

I've not tried a 200MHz Pentium as a server, and I've certainly not tried
one with only 96MB

For reference, my slimserver box is a PIII 500, with 512Mb memory. This is
lower than I'd like, but seems to work just fine.

If the proxy itself is spinning too fast, then try increasing the timeout
value in the script (it's directly below a comment saying 'don't change
unless you know what you're doing *grin*). This will lower the CPU
potentially lower the CPU consumption, but it might make the audio skip if
you go too far.

-- Jules

----- Original Message -----
From: "Triode" <triode1 (AT) btinternet (DOT) com>
To: <discuss (AT) lists (DOT) slimdevices.com>
Sent: Sunday, September 26, 2004 2:10 PM
Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems


> Hi Jules,
>
>> Can you try a test for me? Restart the server, proxy, and kill off any
>> alienstream zombies. Start the proxy in a separate shell, so that you can
>> see it's output (when you first start it, it has nothing to say... just
>> leave it there). Then start slimserver, and play a stream. Try stopping
>> the
>> stream in a bit, and see if the proxy notices and tries to clean up. If
>> it
>> does, check that alienstream isn't still running anyway. If that works,
>> then
>> we'll need to work out why you're getting zombies piling up.
>
> Please see attached for logs - outputs from top and from the proxy.
>
> I should say from the start that my server is a 200 MHz Pentium with 96M
> memory and Mandrake 10. It runs slimserver fine with PCM streaming from
> FLAC etc but stalls with concurrent access to the web interface [I turn
> off
> web refresh]. So I may be stretching this more than others and it may
> just
> be lack of CPU.
>
> My observations:
> 1) CPU load of the transcoder is independant of how many alienstream
> processes are created. Not that high CPU load is seen before any streams
> are started and is highest when nothing else is going on. I wonder if
> this
> is just the interaction of the timer you mentioned and my CPU speed. Any
> suggestion of whether this can be increased? CPU load of this process
> also
> seems to consume any spare cycles upto 70 odd percent..
>
> 2) Although top shows multiple alienstreams, this may be normal? They
> seem
> to start and stop together and most of the CPU time goes against one of
> them?
>
> Will look at mplayer now...
>
> Thanks again,
>
> Adrian
>


--------------------------------------------------------------------------------


>

[Triode]

Jules,

Quick test of chaning the timeout:

Set to 0.1 - cpu load when idle down to 5%
Set to 0.3 - idle cpu load down to <2% (ideal!)

When running a stream, 0.1 seems to work - CPU for transcoder rises to 10%
[22% for main alienstream process]
At 0.3 it gives error messages on output of transcoder_proxy saying it can't
write etc - I assume this is the cause of audio skiping you are discussing?

Is there anyway for the code to have two modes - one when it thinks it is
being used (and hence able to use shorter timeout), and one when idle?

Adrian
----- Original Message -----
From: "Jules Taplin" <slim-discuss (AT) ourhouse (DOT) org.uk>
To: "Slim Devices Discussion" <discuss (AT) lists (DOT) slimdevices.com>
Sent: Sunday, September 26, 2004 3:26 PM
Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems


> Hi Triode.
>
> Just stopped by my machine for 2 mins. I'll try to look at this better
> later.
>
> I've not tried a 200MHz Pentium as a server, and I've certainly not tried
> one with only 96MB
>
> For reference, my slimserver box is a PIII 500, with 512Mb memory. This is
> lower than I'd like, but seems to work just fine.
>
> If the proxy itself is spinning too fast, then try increasing the timeout
> value in the script (it's directly below a comment saying 'don't change
> unless you know what you're doing *grin*). This will lower the CPU
> potentially lower the CPU consumption, but it might make the audio skip if
> you go too far.
>

[Jules Taplin]

Hi Adrian.

I'll have a look. Checking... on my box, it idles at <1% CPU when it's not
in use... so I've never really cared. However... it shouldn't be too tough
to slow it down when there's nothing to service.

Are you running with the lame transcoder enabled, or disabled? The low
timeouts are likely to be necessary when we're pushing raw PCM through the
pipe, but we might get away with them on mp3.

-- Jules

----- Original Message -----
From: "Triode" <triode1 (AT) btinternet (DOT) com>
To: "Slim Devices Discussion" <discuss (AT) lists (DOT) slimdevices.com>
Sent: Sunday, September 26, 2004 4:33 PM
Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems


> Jules,
>
> Quick test of chaning the timeout:
>
> Set to 0.1 - cpu load when idle down to 5%
> Set to 0.3 - idle cpu load down to <2% (ideal!)
>
> When running a stream, 0.1 seems to work - CPU for transcoder rises to 10%
> [22% for main alienstream process]
> At 0.3 it gives error messages on output of transcoder_proxy saying it
> can't write etc - I assume this is the cause of audio skiping you are
> discussing?
>
> Is there anyway for the code to have two modes - one when it thinks it is
> being used (and hence able to use shorter timeout), and one when idle?
>
> Adrian
> ----- Original Message -----
> From: "Jules Taplin" <slim-discuss (AT) ourhouse (DOT) org.uk>
> To: "Slim Devices Discussion" <discuss (AT) lists (DOT) slimdevices.com>
> Sent: Sunday, September 26, 2004 3:26 PM
> Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems
>
>
>> Hi Triode.
>>
>> Just stopped by my machine for 2 mins. I'll try to look at this better
>> later.
>>
>> I've not tried a 200MHz Pentium as a server, and I've certainly not tried
>> one with only 96MB
>>
>> For reference, my slimserver box is a PIII 500, with 512Mb memory. This
>> is lower than I'd like, but seems to work just fine.
>>
>> If the proxy itself is spinning too fast, then try increasing the timeout
>> value in the script (it's directly below a comment saying 'don't change
>> unless you know what you're doing *grin*). This will lower the CPU
>> potentially lower the CPU consumption, but it might make the audio skip
>> if you go too far.
>>
>
>

[Triode]

Hi Jules,

Running with 0.1 and raw PCM seems to work OK- just listened to an entire
episode of hitch-hikers guide on my squeezebox (not softsqueeze) and worked
without drop-outs. However I have noticed how many of the BBC services are
not in 44.1 and hence don't work in this mode - so am going back to look at
lame. Initial view is that I haven't got enough CPU to run lame (running -q
9 and it still needs >30%)

I have tried looking at sox for resampling, but it seems to need you to tell
it what the sampling rate is. Is there any way to get this from the stream
before starting it?

Adrian


----- Original Message -----
From: "Jules Taplin" <slim-discuss (AT) ourhouse (DOT) org.uk>
To: "Slim Devices Discussion" <discuss (AT) lists (DOT) slimdevices.com>
Sent: Sunday, September 26, 2004 6:05 PM
Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems


> Hi Adrian.
>
> I'll have a look. Checking... on my box, it idles at <1% CPU when it's not
> in use... so I've never really cared. However... it shouldn't be too tough
> to slow it down when there's nothing to service.
>
> Are you running with the lame transcoder enabled, or disabled? The low
> timeouts are likely to be necessary when we're pushing raw PCM through the
> pipe, but we might get away with them on mp3.
>
> -- Jules
>
> ----- Original Message -----
> From: "Triode" <triode1 (AT) btinternet (DOT) com>
> To: "Slim Devices Discussion" <discuss (AT) lists (DOT) slimdevices.com>
> Sent: Sunday, September 26, 2004 4:33 PM
> Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems
>
>
>> Jules,
>>
>> Quick test of chaning the timeout:
>>
>> Set to 0.1 - cpu load when idle down to 5%
>> Set to 0.3 - idle cpu load down to <2% (ideal!)
>>
>> When running a stream, 0.1 seems to work - CPU for transcoder rises to
>> 10% [22% for main alienstream process]
>> At 0.3 it gives error messages on output of transcoder_proxy saying it
>> can't write etc - I assume this is the cause of audio skiping you are
>> discussing?
>>
>> Is there anyway for the code to have two modes - one when it thinks it is
>> being used (and hence able to use shorter timeout), and one when idle?
>>
>> Adrian
>> ----- Original Message -----
>> From: "Jules Taplin" <slim-discuss (AT) ourhouse (DOT) org.uk>
>> To: "Slim Devices Discussion" <discuss (AT) lists (DOT) slimdevices.com>
>> Sent: Sunday, September 26, 2004 3:26 PM
>> Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems
>>
>>
>>> Hi Triode.
>>>
>>> Just stopped by my machine for 2 mins. I'll try to look at this better
>>> later.
>>>
>>> I've not tried a 200MHz Pentium as a server, and I've certainly not
>>> tried one with only 96MB
>>>
>>> For reference, my slimserver box is a PIII 500, with 512Mb memory. This
>>> is lower than I'd like, but seems to work just fine.
>>>
>>> If the proxy itself is spinning too fast, then try increasing the
>>> timeout value in the script (it's directly below a comment saying 'don't
>>> change unless you know what you're doing *grin*). This will lower the
>>> CPU potentially lower the CPU consumption, but it might make the audio
>>> skip if you go too far.
>>>
>>
>>

[Jules Taplin]

Hey Adrian.

If you can get the mplayer version to work, then I think we've licked the
resampling issue. mplayer will do it 'in-flight' for us.

In fact... give me 2 minutes, and I'll send you a pre-release of version
0.10 off-list. I've not yet finished a feature or two (mostly interface)
that I was planning, but it's otherwise looking pretty stable, I think.



-- Jules

----- Original Message -----
From: "Triode" <triode1 (AT) btinternet (DOT) com>
To: "Slim Devices Discussion" <discuss (AT) lists (DOT) slimdevices.com>
Sent: Sunday, September 26, 2004 6:34 PM
Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems


> Hi Jules,
>
> Running with 0.1 and raw PCM seems to work OK- just listened to an entire
> episode of hitch-hikers guide on my squeezebox (not softsqueeze) and
> worked without drop-outs. However I have noticed how many of the BBC
> services are not in 44.1 and hence don't work in this mode - so am going
> back to look at lame. Initial view is that I haven't got enough CPU to
> run lame (running -q 9 and it still needs >30%)
>
> I have tried looking at sox for resampling, but it seems to need you to
> tell it what the sampling rate is. Is there any way to get this from the
> stream before starting it?
>
> Adrian
>
>
> ----- Original Message -----
> From: "Jules Taplin" <slim-discuss (AT) ourhouse (DOT) org.uk>
> To: "Slim Devices Discussion" <discuss (AT) lists (DOT) slimdevices.com>
> Sent: Sunday, September 26, 2004 6:05 PM
> Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems
>
>
>> Hi Adrian.
>>
>> I'll have a look. Checking... on my box, it idles at <1% CPU when it's
>> not in use... so I've never really cared. However... it shouldn't be too
>> tough to slow it down when there's nothing to service.
>>
>> Are you running with the lame transcoder enabled, or disabled? The low
>> timeouts are likely to be necessary when we're pushing raw PCM through
>> the pipe, but we might get away with them on mp3.
>>
>> -- Jules
>>
>> ----- Original Message -----
>> From: "Triode" <triode1 (AT) btinternet (DOT) com>
>> To: "Slim Devices Discussion" <discuss (AT) lists (DOT) slimdevices.com>
>> Sent: Sunday, September 26, 2004 4:33 PM
>> Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems
>>
>>
>>> Jules,
>>>
>>> Quick test of chaning the timeout:
>>>
>>> Set to 0.1 - cpu load when idle down to 5%
>>> Set to 0.3 - idle cpu load down to <2% (ideal!)
>>>
>>> When running a stream, 0.1 seems to work - CPU for transcoder rises to
>>> 10% [22% for main alienstream process]
>>> At 0.3 it gives error messages on output of transcoder_proxy saying it
>>> can't write etc - I assume this is the cause of audio skiping you are
>>> discussing?
>>>
>>> Is there anyway for the code to have two modes - one when it thinks it
>>> is being used (and hence able to use shorter timeout), and one when
>>> idle?
>>>
>>> Adrian
>>> ----- Original Message -----
>>> From: "Jules Taplin" <slim-discuss (AT) ourhouse (DOT) org.uk>
>>> To: "Slim Devices Discussion" <discuss (AT) lists (DOT) slimdevices.com>
>>> Sent: Sunday, September 26, 2004 3:26 PM
>>> Subject: [slim] SlimServer 2004_09_23-1 and AlienBBC 0.9 problems
>>>
>>>
>>>> Hi Triode.
>>>>
>>>> Just stopped by my machine for 2 mins. I'll try to look at this better
>>>> later.
>>>>
>>>> I've not tried a 200MHz Pentium as a server, and I've certainly not
>>>> tried one with only 96MB
>>>>
>>>> For reference, my slimserver box is a PIII 500, with 512Mb memory. This
>>>> is lower than I'd like, but seems to work just fine.
>>>>
>>>> If the proxy itself is spinning too fast, then try increasing the
>>>> timeout value in the script (it's directly below a comment saying
>>>> 'don't change unless you know what you're doing *grin*). This will
>>>> lower the CPU potentially lower the CPU consumption, but it might make
>>>> the audio skip if you go too far.
>>>>
>>>
>>>