[mythtv-users] mythcomflag - how do I set nice level?

2005-09-22 Thread Mark Knecht
Hi,
   Is it possible to somehow nice mythcomflag? It appears that in
cases where I'm recording two shows and also doing commercial
detection on two shows that just completed that my poor 3GHz P4 cannot
keep up. This is resulting in bad recordings. I think that if I
lowered priorities on mythcomflag I'd do better.

Thanks,
Mark
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-22 Thread Robert Tsai
On Thu, Sep 22, 2005 at 06:18:11PM -0700, Mark Knecht wrote:
> Is it possible to somehow nice mythcomflag? It appears that in cases
> where I'm recording two shows and also doing commercial detection on
> two shows that just completed that my poor 3GHz P4 cannot keep up.
> This is resulting in bad recordings. I think that if I lowered
> priorities on mythcomflag I'd do better.

In mythtv-setup General options there is a "Job Queue" configuration
page. Set the CPU Usage option:

- High: no nicing
- Medium: nice 19
- Low: nice 19 + sleep() every so often in the main loop

Also, you might want to reduce the number of maximum simultaneous jobs
to "1", unless you have hyperthreading enabled and are running an SMP
kernel.

--Rob


signature.asc
Description: Digital signature
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-22 Thread Mark Knecht
Thanks very much.

I did have it set to one job, but CPU usage was medium so I've set it
to low. Additionally I saw an option to start commercial flagging when
recording starts. I've disabled that also, at least as a test.

Probably 70% of our recording is late night stuff so commercial
flagging can get done between 3AM and 7AM and probably cause no
problems at all.

Again, thanks!

- Mark

On 9/22/05, Robert Tsai <[EMAIL PROTECTED]> wrote:
> On Thu, Sep 22, 2005 at 06:18:11PM -0700, Mark Knecht wrote:
> > Is it possible to somehow nice mythcomflag? It appears that in cases
> > where I'm recording two shows and also doing commercial detection on
> > two shows that just completed that my poor 3GHz P4 cannot keep up.
> > This is resulting in bad recordings. I think that if I lowered
> > priorities on mythcomflag I'd do better.
>
> In mythtv-setup General options there is a "Job Queue" configuration
> page. Set the CPU Usage option:
>
> - High: no nicing
> - Medium: nice 19
> - Low: nice 19 + sleep() every so often in the main loop
>
> Also, you might want to reduce the number of maximum simultaneous jobs
> to "1", unless you have hyperthreading enabled and are running an SMP
> kernel.
>
> --Rob
>
>
> -BEGIN PGP SIGNATURE-
> Version: GnuPG v1.4.1 (GNU/Linux)
>
> iD8DBQFDM1v+18A6pRrdKfkRAur6AJ0VUMTbhBrA1Ucbu7ZDwiI2kJnJpACfUUx0
> OSiN3NV6bLgLDg2PohSNrlQ=
> =vRBs
> -END PGP SIGNATURE-
>
>
> ___
> mythtv-users mailing list
> mythtv-users@mythtv.org
> http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
>
>
>
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-22 Thread Robert Tsai
On Thu, Sep 22, 2005 at 06:52:15PM -0700, Mark Knecht wrote:
> Thanks very much.
> 
> I did have it set to one job, but CPU usage was medium so I've set it
> to low. Additionally I saw an option to start commercial flagging when
> recording starts. I've disabled that also, at least as a test.
> 
> Probably 70% of our recording is late night stuff so commercial
> flagging can get done between 3AM and 7AM and probably cause no
> problems at all.

I think your real problem was commercial flagging when recording
starts (a.k.a. "real-time commercial flagging).

You might be OK bumping CPU usage back up to medium so that
commflagging can go faster. The nicing should prevent mythcommflagging
from interfering with recordings. But I don't know the details of your
capture cards, maybe you need more CPU for software encoding.

My AMD64 always runs in powersave mode at 1GHz and lets me watch one
recording, record two more off of HD-3000 cards (digital broadcast so
no encoding), and mythcommflagging, all at the same time, with no
problems with anything.

> On 9/22/05, Robert Tsai <[EMAIL PROTECTED]> wrote:
> > On Thu, Sep 22, 2005 at 06:18:11PM -0700, Mark Knecht wrote:
> > > Is it possible to somehow nice mythcomflag? It appears that in
> > > cases where I'm recording two shows and also doing commercial
> > > detection on two shows that just completed that my poor 3GHz P4
> > > cannot keep up. This is resulting in bad recordings. I think
> > > that if I lowered priorities on mythcomflag I'd do better.
> >
> > In mythtv-setup General options there is a "Job Queue"
> > configuration page. Set the CPU Usage option:
> >
> > - High: no nicing
> > - Medium: nice 19
> > - Low: nice 19 + sleep() every so often in the main loop
> >
> > Also, you might want to reduce the number of maximum simultaneous
> > jobs to "1", unless you have hyperthreading enabled and are
> > running an SMP kernel.

--Rob


signature.asc
Description: Digital signature
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-22 Thread Mark Knecht
On 9/22/05, Robert Tsai <[EMAIL PROTECTED]> wrote:
> On Thu, Sep 22, 2005 at 06:52:15PM -0700, Mark Knecht wrote:
> > Thanks very much.
> >
> > I did have it set to one job, but CPU usage was medium so I've set it
> > to low. Additionally I saw an option to start commercial flagging when
> > recording starts. I've disabled that also, at least as a test.
> >
> > Probably 70% of our recording is late night stuff so commercial
> > flagging can get done between 3AM and 7AM and probably cause no
> > problems at all.
>
> I think your real problem was commercial flagging when recording
> starts (a.k.a. "real-time commercial flagging).

Could be. Now that I know where the setting are I can play with that a bit.
>
> You might be OK bumping CPU usage back up to medium so that
> commflagging can go faster. The nicing should prevent mythcommflagging
> from interfering with recordings. But I don't know the details of your
> capture cards, maybe you need more CPU for software encoding.

3GHz P4HT but currently a UMP kernel.
PVR-150
PVR-250
512MB

250GB storage over NFS in another box so good networking matters.

>
> My AMD64 always runs in powersave mode at 1GHz and lets me watch one
> recording, record two more off of HD-3000 cards (digital broadcast so
> no encoding), and mythcommflagging, all at the same time, with no
> problems with anything.

Sounds cool.

>
> > On 9/22/05, Robert Tsai <[EMAIL PROTECTED]> wrote:
> > > On Thu, Sep 22, 2005 at 06:18:11PM -0700, Mark Knecht wrote:
> > > > Is it possible to somehow nice mythcomflag? It appears that in
> > > > cases where I'm recording two shows and also doing commercial
> > > > detection on two shows that just completed that my poor 3GHz P4
> > > > cannot keep up. This is resulting in bad recordings. I think
> > > > that if I lowered priorities on mythcomflag I'd do better.
> > >
> > > In mythtv-setup General options there is a "Job Queue"
> > > configuration page. Set the CPU Usage option:
> > >
> > > - High: no nicing
> > > - Medium: nice 19
> > > - Low: nice 19 + sleep() every so often in the main loop
> > >
> > > Also, you might want to reduce the number of maximum simultaneous
> > > jobs to "1", unless you have hyperthreading enabled and are
> > > running an SMP kernel.
>
> --Rob
>
>
> -BEGIN PGP SIGNATURE-
> Version: GnuPG v1.4.1 (GNU/Linux)
>
> iD8DBQFDM2ID18A6pRrdKfkRAsu7AKCSy4gJlfbAxl4ouNNQ4R9AMIw4swCfcw4S
> /0W456aC2dINUvPn2Iv4aUI=
> =yW8V
> -END PGP SIGNATURE-
>
>
> ___
> mythtv-users mailing list
> mythtv-users@mythtv.org
> http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
>
>
>
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-23 Thread Mark Knecht
On 9/22/05, Mark Knecht <[EMAIL PROTECTED]> wrote:
> On 9/22/05, Robert Tsai <[EMAIL PROTECTED]> wrote:
> > On Thu, Sep 22, 2005 at 06:52:15PM -0700, Mark Knecht wrote:
> > > Thanks very much.
> > >
> > > I did have it set to one job, but CPU usage was medium so I've set it
> > > to low. Additionally I saw an option to start commercial flagging when
> > > recording starts. I've disabled that also, at least as a test.
> > >
> > > Probably 70% of our recording is late night stuff so commercial
> > > flagging can get done between 3AM and 7AM and probably cause no
> > > problems at all.
> >
> > I think your real problem was commercial flagging when recording
> > starts (a.k.a. "real-time commercial flagging).
>
> Could be. Now that I know where the setting are I can play with that a bit.
> >
> > You might be OK bumping CPU usage back up to medium so that
> > commflagging can go faster. The nicing should prevent mythcommflagging
> > from interfering with recordings. But I don't know the details of your
> > capture cards, maybe you need more CPU for software encoding.
>
> 3GHz P4HT but currently a UMP kernel.
> PVR-150
> PVR-250
> 512MB
>
> 250GB storage over NFS in another box so good networking matters.

Hi Robert,
   Well, last night's recordings mostly failed again:

8PM - 1 30 minute show - recorded fine. Recording info says 30 minutes long

10PM - 2 30 minute shows. Both failed. One is 18 minutes, the other is
19 minutes.

11PM - 1 30 minute show - looks good

11:30PM - 2 60 minute shows - both failed - One is 22 minutes, the
other is 25 minutes

12:30PM - 1 60 minute show - failed. Info says 37 minutes.

   Looks like I've got more problems than just the settings we played
with yesterday. However, a question - Did I need to restart the
backend server to get those changes to take effect? I didn't do that.

   I'll try that this morning as a test.

   Here's the failure messages in dmesg"

ivtv: ENC Stream 0 OVERFLOW #57: Stealing a Buffer, 512 currently allocated
ivtv: ENC Stream 0 OVERFLOW #58: Stealing a Buffer, 512 currently allocated
ivtv: ENC Stream 0 OVERFLOW #59: Stealing a Buffer, 512 currently allocated
ivtv: ENC Stream 0 OVERFLOW #60: Stealing a Buffer, 512 currently allocated
ivtv: ENC Stream 0 OVERFLOW #61: Stealing a Buffer, 512 currently allocated
ivtv: ENC Stream 0 OVERFLOW #62: Stealing a Buffer, 512 currently allocated
ivtv: ENC Stream 0 OVERFLOW #63: Stealing a Buffer, 512 currently allocated
[EMAIL PROTECTED] ~ $

Thanks,
Mark
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-23 Thread Ryan Steffes
On 9/23/05, Mark Knecht <[EMAIL PROTECTED]> wrote:
On 9/22/05, Mark Knecht <[EMAIL PROTECTED]> wrote:> On 9/22/05, Robert Tsai <[EMAIL PROTECTED]> wrote:> > On Thu, Sep 22, 2005 at 06:52:15PM -0700, Mark Knecht wrote:
> > > Thanks very much.> > >> > > I did have it set to one job, but CPU usage was medium so I've set it> > > to low. Additionally I saw an option to start commercial flagging when
> > > recording starts. I've disabled that also, at least as a test.> > >> > > Probably 70% of our recording is late night stuff so commercial> > > flagging can get done between 3AM and 7AM and probably cause no
> > > problems at all.> >> > I think your real problem was commercial flagging when recording> > starts (a.k.a. "real-time commercial flagging).>> Could be. Now that I know where the setting are I can play with that a bit.
> >> > You might be OK bumping CPU usage back up to medium so that> > commflagging can go faster. The nicing should prevent mythcommflagging> > from interfering with recordings. But I don't know the details of your
> > capture cards, maybe you need more CPU for software encoding.>> 3GHz P4HT but currently a UMP kernel.> PVR-150> PVR-250> 512MB>> 250GB storage over NFS in another box so good networking matters.
Hi Robert,   Well, last night's recordings mostly failed again:8PM - 1 30 minute show - recorded fine. Recording info says 30 minutes long10PM - 2 30 minute shows. Both failed. One is 18 minutes, the other is
19 minutes.11PM - 1 30 minute show - looks good11:30PM - 2 60 minute shows - both failed - One is 22 minutes, theother is 25 minutes12:30PM - 1 60 minute show - failed. Info says 37 minutes.
   Looks like I've got more problems than just the settings we playedwith yesterday. However, a question - Did I need to restart thebackend server to get those changes to take effect? I didn't do that.
   I'll try that this morning as a test.   Here's the failure messages in dmesg"ivtv: ENC Stream 0 OVERFLOW #57: Stealing a Buffer, 512 currently allocatedivtv: ENC Stream 0 OVERFLOW #58: Stealing a Buffer, 512 currently allocated
ivtv: ENC Stream 0 OVERFLOW #59: Stealing a Buffer, 512 currently allocatedivtv: ENC Stream 0 OVERFLOW #60: Stealing a Buffer, 512 currently allocatedivtv: ENC Stream 0 OVERFLOW #61: Stealing a Buffer, 512 currently allocated
ivtv: ENC Stream 0 OVERFLOW #62: Stealing a Buffer, 512 currently allocatedivtv: ENC Stream 0 OVERFLOW #63: Stealing a Buffer, 512 currently allocated[EMAIL PROTECTED] ~ $

You've very likely got other problems if you are having those problems
with that set up.  Have you checked for things like IRQ conflicts,
and made sure you are using a current IVTV that's configured correctly,
made sure DMA is enable on your drives, that sort of thing?  With
the two PVRs, your CPU really shouldn't enter much into it as far as
recording goes.  Your problem is likely to be elsewhere.  

___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-23 Thread Mark Knecht
On 9/23/05, Ryan Steffes <[EMAIL PROTECTED]> wrote:
>
>
>
> On 9/23/05, Mark Knecht <[EMAIL PROTECTED]> wrote:

> > ivtv: ENC Stream 0 OVERFLOW #61: Stealing a Buffer, 512 currently
> allocated
> > ivtv: ENC Stream 0 OVERFLOW #62: Stealing a Buffer, 512 currently
> allocated
> > ivtv: ENC Stream 0 OVERFLOW #63: Stealing a Buffer, 512 currently
> allocated
> > [EMAIL PROTECTED] ~ $
>
>
>  You've very likely got other problems if you are having those problems with
> that set up.  Have you checked for things like IRQ conflicts, and made sure
> you are using a current IVTV that's configured correctly, made sure DMA is
> enable on your drives, that sort of thing?  With the two PVRs, your CPU
> really shouldn't enter much into it as far as recording goes.  Your problem
> is likely to be elsewhere.
>

Hi,
   Yes. DMA is running on the local hard drives but the storage for
MythTV is over the network on an NFS mount. Strange thing is that it's
worked for 2 months just fine and only started failing 2 days ago.

   There is not way, that I know of, to make sure ivtv is 'configured
correctly'. There are no error messages when the drivers are loaded,
but that driver prints so many messages it looks like black magic. ;-)

   The sort of strange thing I've noticed with this new problem is
that when recording, if I run top, I see almost 100% of time spent
'waiting'. I am told on other lists that this is a measure of DMA
taking place. The CPU has issued commands and is now waiting for them
to complete. That would be good if the machine wasn't failing. I do
not know how to tell if the machine is 'waiting' for the two PVR cards
or 'waiting' for the NIC to transfer and store data.

Thanks,
Mark
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-23 Thread Ryan Steffes

>They do not appear to be.ivtv0 = 22ivtv1 = 17eth0 = 20
However ivtv1 is sharing with some part fo the chipset:dragonfly ~ # cat /proc/interrupts   CPU0  0:   14871236IO-APIC-edge  timer  1:  23821IO-APIC-edge  i8042  9:  0   IO-APIC-level  acpi
 12: 308115IO-APIC-edge  i8042 14: 187334IO-APIC-edge  ide0 15: 49IO-APIC-edge  ide1 16: 485287   IO-APIC-level  uhci_hcd:usb2, uhci_hcd:usb5,[EMAIL PROTECTED]::00:02.0 17:  57811   IO-APIC-level  ivtv1, Intel ICH5
 18:109   IO-APIC-level  uhci_hcd:usb4 19:174   IO-APIC-level  uhci_hcd:usb3 20:8034881   IO-APIC-level  eth0 22:  38977   IO-APIC-level  ivtv0 23:  4   IO-APIC-level  ehci_hcd:usb1
NMI:  0LOC:   14872124ERR:  0MIS:  0dragonfly ~ #

I was having a very similiar problem, which is why I asked.  For
me, it wasn't just failing though it was resulting in lock ups. 
The solution was to turn off APIC in the BIOS and it's been hunky dory
since.
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-23 Thread Mark Knecht
On 9/23/05, Ryan Steffes <[EMAIL PROTECTED]> wrote:
>
>
> > >
> > They do not appear to be.
> >
> > ivtv0 = 22
> > ivtv1 = 17
> > eth0 = 20
> >
> > However ivtv1 is sharing with some part fo the chipset:
> >
> > dragonfly ~ # cat /proc/interrupts
> >CPU0
> >   0:   14871236IO-APIC-edge  timer
> >   1:  23821IO-APIC-edge  i8042
> >   9:  0   IO-APIC-level  acpi
> > 12: 308115IO-APIC-edge  i8042
> > 14: 187334IO-APIC-edge  ide0
> > 15: 49IO-APIC-edge  ide1
> > 16: 485287   IO-APIC-level  uhci_hcd:usb2, uhci_hcd:usb5,
> > [EMAIL PROTECTED]::00:02.0
> > 17:  57811   IO-APIC-level  ivtv1, Intel ICH5
> > 18:109   IO-APIC-level  uhci_hcd:usb4
> > 19:174   IO-APIC-level  uhci_hcd:usb3
> > 20:8034881   IO-APIC-level  eth0
> > 22:  38977   IO-APIC-level  ivtv0
> > 23:  4   IO-APIC-level  ehci_hcd:usb1
> > NMI:  0
> > LOC:   14872124
> > ERR:  0
> > MIS:  0
> > dragonfly ~ #
> >
>
>
>  I was having a very similiar problem, which is why I asked.  For me, it
> wasn't just failing though it was resulting in lock ups.  The solution was
> to turn off APIC in the BIOS and it's been hunky dory since.
>

As a test I jsut switched video storage to my local drive and ran a 30
minute test run recording two programs. Both programs came out 29:55
second, basically perfect.

One difference I noted was while running this test top said that I was
98% idle. Sounds great. While running an identical 30 minute test
storing on my NFS drive top said I was 98% 'waiting'.

I think my problem resides somewhere in the NFS system. I also note
looking at the output of ifconfig that both the backend side and the
NFS server are logging thousand up thousand 'collisions'. I need to
fix that next.

Thanks,
Mark
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-23 Thread Roman Romaniuk
>On 9/23/05, Ryan Steffes <[EMAIL PROTECTED]> wrote:
>>
>>
>> > >
>> > They do not appear to be.
>> >
>> > ivtv0 = 22
>> > ivtv1 = 17
>> > eth0 = 20
>> >
>> > However ivtv1 is sharing with some part fo the chipset:
>> >
>> > dragonfly ~ # cat /proc/interrupts
>> >CPU0
>> >   0:   14871236IO-APIC-edge  timer
>> >   1:  23821IO-APIC-edge  i8042
>> >   9:  0   IO-APIC-level  acpi
>> > 12: 308115IO-APIC-edge  i8042
>> > 14: 187334IO-APIC-edge  ide0
>> > 15: 49IO-APIC-edge  ide1
>> > 16: 485287   IO-APIC-level  uhci_hcd:usb2, uhci_hcd:usb5,
>> > [EMAIL PROTECTED]::00:02.0
>> > 17:  57811   IO-APIC-level  ivtv1, Intel ICH5
>> > 18:109   IO-APIC-level  uhci_hcd:usb4
>> > 19:174   IO-APIC-level  uhci_hcd:usb3
>> > 20:8034881   IO-APIC-level  eth0
>> > 22:  38977   IO-APIC-level  ivtv0
>> > 23:  4   IO-APIC-level  ehci_hcd:usb1
>> > NMI:  0
>> > LOC:   14872124
>> > ERR:  0
>> > MIS:  0
>> > dragonfly ~ #
>> >
>>
>>
>>  I was having a very similiar problem, which is why I asked.  For me, it
>> wasn't just failing though it was resulting in lock ups.  The solution
wa>s
>> to turn off APIC in the BIOS and it's been hunky dory since.
>>
>
>As a test I jsut switched video storage to my local drive and ran a 30
>minute test run recording two programs. Both programs came out 29:55
>second, basically perfect.
>
>One difference I noted was while running this test top said that I was
>98% idle. Sounds great. While running an identical 30 minute test
>storing on my NFS drive top said I was 98% 'waiting'.
>
>I think my problem resides somewhere in the NFS system. I also note
>looking at the output of ifconfig that both the backend side and the
>NFS server are logging thousand up thousand 'collisions'. I need to
>fix that next.
>
>Thanks,
>Mark
>___
>mythtv-users mailing list
>mythtv-users@mythtv.org
>http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
>
>

NFS works very well for reading, not so well for writing.  NFS absolutely
insists (for a whole bunch of valid reasons) on commiting the data to disk
before returning,  resulting in very poor write performance as there is no
buffering. Your best bet is writing to a local hard disk.

Cheers,

Roman

Sent using cyberus.ca WebMail - http://www.cyberus.ca/
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-23 Thread Mark Cooke
On Fri, 2005-09-23 at 15:55 -0500, Roman Romaniuk wrote:
> >On 9/23/05, Ryan Steffes <[EMAIL PROTECTED]> wrote:
> >
> >I think my problem resides somewhere in the NFS system. I also note
> >looking at the output of ifconfig that both the backend side and the
> >NFS server are logging thousand up thousand 'collisions'. I need to
> >fix that next.

This sounds like a half vs full duplex problem.

> NFS works very well for reading, not so well for writing.  NFS absolutely
> insists (for a whole bunch of valid reasons) on commiting the data to disk
> before returning,  resulting in very poor write performance as there is no
> buffering. Your best bet is writing to a local hard disk.

In the server side /etc/exports file I have:

# TV system exports
/mnt/d4/tv   client(rw,no_root_squash,async,no_subtree_check)

The async helps with the write performance but at the risk of data loss
if the power goes down.

I also have no version 2 support, and use nfs over tcp rather than udp.


On the client side fstab:

server:/mnt/d4/tv /mnt/d4/tv nfs  \
defaults,wsize=32768,rsize=8192,noatime,hard,nfsvers=3,async 1 1


Provides good performance for my setup (3 streams over NFS without a
hiccup).

YMMV.

Cheers,

Mark


___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-09-24 Thread Mark Knecht
On 9/23/05, Mark Cooke <[EMAIL PROTECTED]> wrote:
> On Fri, 2005-09-23 at 15:55 -0500, Roman Romaniuk wrote:
> > >On 9/23/05, Ryan Steffes <[EMAIL PROTECTED]> wrote:
> > >
> > >I think my problem resides somewhere in the NFS system. I also note
> > >looking at the output of ifconfig that both the backend side and the
> > >NFS server are logging thousand up thousand 'collisions'. I need to
> > >fix that next.
>
> This sounds like a half vs full duplex problem.
>

In the end that's exactly what it was. We apparantly had a small power
drop on Tuesday. I could see in one of my logs that a Pundit-R had
rebooted as it doesn't have a UPS. My switch runs off the same wall
plug. Apparently the switch got into a strange state and would no
longer negotiate Ethernet speeds and had gone all the way down to 10Mb
- Half Duplex. Unfortunately the switch was hidden from sight so I
couldn't see the lights on the front panel.

Once I pulled the power on the switch and essentially rebooted it I
went back to 100BaseTx-Fulle Duplex and now have no problems with Myth
at all.

thanks for the info on your setup. I'll look into whether it would
improve my setup also. I currently do not run async but probably
could.

Cheers,
Mark
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users


Re: [mythtv-users] mythcomflag - how do I set nice level?

2005-10-07 Thread Tim Sailer
On Fri, Sep 23, 2005 at 11:40:00AM -0700, Mark Knecht wrote:
> As a test I jsut switched video storage to my local drive and ran a 30
> minute test run recording two programs. Both programs came out 29:55
> second, basically perfect.
> 
> One difference I noted was while running this test top said that I was
> 98% idle. Sounds great. While running an identical 30 minute test
> storing on my NFS drive top said I was 98% 'waiting'.
> 
> I think my problem resides somewhere in the NFS system. I also note
> looking at the output of ifconfig that both the backend side and the
> NFS server are logging thousand up thousand 'collisions'. I need to
> fix that next.

Use NFS3, and play with increased rsize and wsize blocks. That makes
a big difference, *as long* as you have a clean full-duplex network.

Tim

-- 
 
 >> Tim Sailer  ><  Coastal Internet, Inc. <<
 >> Network and Systems Operations  ><  PO Box 726 <<
 >> http://www.buoy.com ><  Moriches, NY 11955 <<
 >> [EMAIL PROTECTED]><  (631) 399-2910  (888) 924-3728 <<
 
___
mythtv-users mailing list
mythtv-users@mythtv.org
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users