FOOD FOR THOUGHT Best Timing Source

luckman212

Guru
Joined
Jul 7, 2010
Messages
272
Reaction score
0
I'm now toying with the prospect of completely blowing out this PIAF and reloading 1.6 on it. Just hate going backwards and setting everything up again, especially now with IPv6 on the way. I am totally stumped as to why MOH doesn't work (the standard MOH does, just not streaming MOH) but I have no idea why the dahdi timer would affect only streaming MOH.
 

dswartz

Guru
Joined
Feb 17, 2009
Messages
1,056
Reaction score
0
I'm waiting for ward to post his favorite picture of the guy running away with a flock of arrows chasing him :) I just ordered a UT51 - we'll see...
 

elihunter

New Member
Joined
Jun 13, 2010
Messages
15
Reaction score
0
I'm running Asterisk 1.4.38 and "module show like timing" shows no modules loaded. Is that because I'm on 1.4.38?

I was able to enable HPET in the Bios on my server which seemed to bump the average dahdi_test score up a little but I'm still getting 99.96xx's in during the test, just less frequent. The client did say the calls are clearer and I think there was a pretty noticeable reduction in static.

I ordered a UT50 and am eagerly awaiting testing it.
 

luckman212

Guru
Joined
Jul 7, 2010
Messages
272
Reaction score
0
More interesting-ness... :smile5:
Okay so I believe I've got my system using the DAHDI timing hardware now for asterisk (Sangoma UT50). Here's this:

Code:
[FONT=Fixedsys]pbx*CLI> module show like timing
Module                         Description                              Use Count 
res_timing_dahdi.so            DAHDI Timing Interface                   1         
res_timing_pthread.so          pthread Timing Interface                 1         
2 modules loaded[/FONT]
You'll note that something is still hooking into pthread. I believe it's res_musiconhold.c. I get tons of complaints in /var/log/asterisk/full if I try to spin up asterisk with pthread timing disabled (noload=>res_timing_pthread.so). res_musiconhold.c is littered with references to pthread, so I think it needs it enabled. Perhaps Digium did this to avoid the DAHDI timer requirements of yore (they rewrote MeetMe afaik as well for that reason...) Still not sure how to see which asterisk apps are using which timing modules (anyone know how to do that?)

I made one additional "nice" discovery. Streaming MOH was previously NOT working with res_timing_dahdi enabled. I played around with the /etc/asterisk/modules.conf a bit and despite the suggestions of others to make this change:

noload => res_timing_dahdi.so

I instead found that if I commented out those explicit load= commands in the "Ensure that format_* modules are loaded before res_musiconhold" section:

;load => format_mp3.so
;load => res_musiconhold.so


Instead I just allow the modules to load in the order that they normally would, and I can now successfully load DAHDI and also still have working streaming MOH. Hoo-rah! :D
This is with ast 1.8.2-rc1.
 

kflorian

Member
Joined
Feb 18, 2010
Messages
76
Reaction score
4
Jake,

Can you give a pointer toward instructions on loading the CentOS VM Kernel? I' a *nix noob.

Thanks,

Ken
 

kflorian

Member
Joined
Feb 18, 2010
Messages
76
Reaction score
4
linetux,

Did the ESX test have a Sangoma on it?

I presume not but just checking (as I learn).

Ken
 

dswartz

Guru
Joined
Feb 17, 2009
Messages
1,056
Reaction score
0
well that sucks. i plug in my ut51 that came today. download and install the wanpipe-voicetime driver. fire up dahdi, and... kernel OOPS! Some sort of botch trying to suspend with a lock. I found (no thanks to sangoma docs) that i needed the wanpipe_voicetime code. I pulled down the latest (1.0.10) and built and installed it. Fire up dahdi and it craps out :( Time to open a ticket with sangoma I guess...
 

Linetux

Guru
Joined
Oct 5, 2008
Messages
541
Reaction score
1
No, my ESX installs do not have any timing sources associated with them.
 

dswartz

Guru
Joined
Feb 17, 2009
Messages
1,056
Reaction score
0
Really not my day :( I decided to try an experiment: set up a piaf kvm, and export the ut51 to it. Good news: it finds it and the wanpipe_voicetime stuff builds, installs and loads just fine. Bad news: the results are total crap. It consistently is telling me:

8192 samples in 24718.969 system clock sample intervals (301.745%)
8192 samples in 24661.062 system clock sample intervals (301.038%)
8192 samples in 25303.320 system clock sample intervals (308.878%)
8192 samples in 24686.680 system clock sample intervals (301.351%)

I guess I wait for sangoma to get back to me as to why their driver crashes on a debian 2.6.32 kernel :(
 

dswartz

Guru
Joined
Feb 17, 2009
Messages
1,056
Reaction score
0
Hmmm, one thing I thought was suspicious: although the oops happens immediately on loading the wanpipe_voicetime module, the message is always about:

Pid: 0, comm: swapper Tainted: G W 2.6.32-4-pve #1 dzhanibekov MCP55S

The MCP55S is for this mobo's onboard chipset. Ethernet + SATA. I wonder if they are conflicting somehow?
 

dswartz

Guru
Joined
Feb 17, 2009
Messages
1,056
Reaction score
0
Hmmm, no, I haven't. I guess that is next, although I am not sure what I can actually do if there is one. Everything that can cause a conflict is built-in to the nvidia chipset (the two nics, the usb hubs, the sata controllers, etc...) except for the RAID controller, and there was never an issue before. I didn't think specific USB devices had their own interrupts - do they?
 

randy7376

Defnyddiwr Gweithredol
Joined
Sep 29, 2010
Messages
864
Reaction score
144
If I'm not mistaken, cat /proc/interrupts will show you what you want to know about IRQ usage on your system. On my system, for example:

Code:
           CPU0       CPU1       CPU2       CPU3       
  0:         44          0          1        158   IO-APIC-edge      timer
  1:          2          0          0          2   IO-APIC-edge      i8042
  8:          0          0          0          1   IO-APIC-edge      rtc0
  9:          0          0          0          0   IO-APIC-fasteoi   acpi
 12:          3          0          0          4   IO-APIC-edge      i8042
 14:          0          0          0          0   IO-APIC-edge      pata_atiixp
 15:          0          0          0          0   IO-APIC-edge      pata_atiixp
 16:     104655          7       1825     806940   IO-APIC-fasteoi   ohci_hcd:usb3, ohci_hcd:usb4, hda_intel
 17:          3          0          0          3   IO-APIC-fasteoi   ehci_hcd:usb1
 18:      11498          5        379     289801   IO-APIC-fasteoi   ohci_hcd:usb5, ohci_hcd:usb6, ohci_hcd:usb7, nvidia
 19:     433678         48       9079    3469402   IO-APIC-fasteoi   ehci_hcd:usb2
 22:     199612         40       8270    1692894   IO-APIC-fasteoi   ahci
 42:          2         27       5493    1946479   PCI-MSI-edge      eth0
NMI:          0          0          0          0   Non-maskable interrupts
LOC:    7927596    6877326    9848501    4576224   Local timer interrupts
SPU:          0          0          0          0   Spurious interrupts
PMI:          0          0          0          0   Performance monitoring interrupts
PND:          0          0          0          0   Performance pending work
RES:   19663139   11964604    6456802    4761746   Rescheduling interrupts
CAL:     970041    1044460     923675     537126   Function call interrupts
TLB:     144658      93476      88261      64786   TLB shootdowns
TRM:          0          0          0          0   Thermal event interrupts
THR:          0          0          0          0   Threshold APIC interrupts
MCE:          0          0          0          0   Machine check exceptions
MCP:        177        176        176        176   Machine check polls
ERR:          0
MIS:          0
Also, lspci -v will display IRQ usage (along with lots of additional info) for your PCI devices.
 

dswartz

Guru
Joined
Feb 17, 2009
Messages
1,056
Reaction score
0
here's what has me questioning this: the only add-on device is the PCI 3ware controller, which is not sharing an IRQ with anything. I even tried disabling the two NICs and the 3 sata ports in the bios and loading the wanpipe_voicetime KO and no difference. As far as I can see, a usb device plugged in will not get an IRQ of its own, so I don't see how this can be the cause (and since there is basically nothing added on, what I could change or remove to fix it if it were :( )
 

randy7376

Defnyddiwr Gweithredol
Joined
Sep 29, 2010
Messages
864
Reaction score
144
The USB device itself doesn't get the actual interrupt, but the controller for that port does. In the example posted earlier, ohci_hcd and ehci_hcd show which interrupts are currently in-use.

Take a look at lsusb -v , no actual interrupts are used. A transfer type of interrupt is shown, but that request is handled by the USB controller(s).

If you're not seeing any differences when you disable things in the BIOS, do you have a setting that shows something along the lines of "Plug N' Play OS" yes | no ?

I typically set that option to "no" when running Linux.
 

dswartz

Guru
Joined
Feb 17, 2009
Messages
1,056
Reaction score
0
Hmmm, I will check that, thanks. I don't understand why the usb controller would be involved, since I do have an APC UPS plugged into a USB port, but maybe a different IRQ for that controller. I will check that (and the bios setting...)
 

dswartz

Guru
Joined
Feb 17, 2009
Messages
1,056
Reaction score
0
I think there is something badly broken here. What I am seeing is a message about "scheduling while atomic", which seems to entail going through the scheduler while holding a spinlock. And it happens the moment I load the sangoma driver. I don't know if it is a dahdi bug that only happens with my kernel (2.6.32) or what. given that i have had no complaints about voice quality and given that the results when i hit this bug requires a hard reset on the server, i think i'm going to abandon this. i'll probably be putting the ut51 up for sale here :(
 

dswartz

Guru
Joined
Feb 17, 2009
Messages
1,056
Reaction score
0
I have a low-power atom unit I've had sitting around doing nothing. I may try throwing piaf on it and moving the ut51 to it and seeing what happens...
 

dswartz

Guru
Joined
Feb 17, 2009
Messages
1,056
Reaction score
0
sangoma has asked for all the specific info from me (distro, release, 32 vs 64, etc) and are going to try to repro this.
 

Members online

No members online now.

Forum statistics

Threads
25,781
Messages
167,507
Members
19,201
Latest member
troutpocket
Get 3CX - Absolutely Free!

Link up your team and customers Phone System Live Chat Video Conferencing

Hosted or Self-managed. Up to 10 users free forever. No credit card. Try risk free.

3CX
A 3CX Account with that email already exists. You will be redirected to the Customer Portal to sign in or reset your password if you've forgotten it.
Top