Anthony, Just so you didn't get me wrong (feels like I got stuck in the middle here...). I Just wanted to point out that I do have the same "problem" as Wasim. I'm well aware I'm running this on old hardware, I just use it for lab purposes, so it really doesn't matter. I don't have any audio problems, since I just use it for 1-2 silutaneous calls. But I do get the warning when I start up FS, the same on both CentOS 5.4 and 5.5 (didn't try 5.3 on this box, I started with 5.4 and upgraded yesterday), so I just wanted to try helping out if anyone needed a 5.5 box to try on. I'm running "production" (only for internal use, but still) servers on both CentOS 5.3 and Windows 2003, which work perfectly fine (no timer gap warnings) - but they have good hardware though :) Regards, Peter Från: freeswitch-users-bounces@lists.freeswitch.org [mailto:freeswitch-users-bounces@lists.freeswitch.org] För Anthony Minessale Skickat: den 18 maj 2010 17:35 Till: freeswitch-users@lists.freeswitch.org Ämne: Re: [Freeswitch-users] centos 5.4 timer issue 1000hz, choppy voice at times Please heed as this is our official policy. There is not an issue with CentOS 5.3. In fact CentOS 5.3 is the preferred platform (not 5.4 or 5.5 until we do proper testing and QA). What you are describing is the result of poor hardware, (most likely the motherboard) FreeSWITCH is Free, your box is not, use the money you save not buying FreeSWITCH to buy a decent box. Even a $500 box can work perfectly fine. Its not my intent to cater to inferior hardware when my goal is to push telephony into the future not get caught up in legacy drama with older boxes. If you have altered the kernel to run at something besides 1000hz, change it back to the default of 1000hz. If you get the error instantly it means the kernel is actually configured at the wrong freq. If the test runs for a long time and concludes it, it most likely means the box is not good enough to run FS. If you are not careful you may have installed the xen build of the kernel (check with uname -a) This can also cause timing issues. If you see that timing gap warning (especially on centos if you have it set to 1000hz) it means the motherboard cannot give the OS a good enough timing source to provide 1ms resolution which is mandatory and not very demanding. I have been on dozens of perfectly happy boxes running CentOS 5.3 and I have been on a small handful of boxes that do what you describe and have seen that, once they are where they belong, (in the trash) the new inexpensive replacement does just fine. On Tue, May 18, 2010 at 2:58 AM, Peter Olsson <peter.olsson@visionutveckling.se<mailto:peter.olsson@visionutveckling.se>> wrote: I also have this behaviour (timer gap around 2000). Upgraded yesterday to CentOS 5.5 (from 5.4), but the issue still remains. The system is just used for lab purposes, so I could easily fix SSH access if someone wanted to have a closer look. I should also add - this is a 3-4 year old machine (workstation), but it seems this problem is related to newer CentOS-kernels, so I guess it shouldn't be the issue here. /Peter Från: freeswitch-users-bounces@lists.freeswitch.org<mailto:freeswitch-users-bounces@lists.freeswitch.org> [mailto:freeswitch-users-bounces@lists.freeswitch.org<mailto:freeswitch-users-bounces@lists.freeswitch.org>] För Wasim Baig Skickat: den 18 maj 2010 06:20 Till: FreeSWITCH Users Ämne: [Freeswitch-users] centos 5.4 timer issue 1000hz, choppy voice at times I know its been discussed previously on list, but has there been any progress to the timer issue on centos 5.4 32 bit Linux pum2 2.6.18-164.el5 #1 SMP Thu Sep 3 03:33:56 EDT 2009 i686 i686 i386 GNU/Linux [root@pum2 ~]# grep 1000 /boot/config-2.6.18-164.el5 CONFIG_HZ_1000=y CONFIG_HZ=1000 i used a basic timer testing utility to confirm what we're actually getting [root@pum2 ~]# ./timerchk kernel timer interrupt frequency is approx. 1002 Hz however, fs complains ... 2010-05-18 06:17:24.711360 [CONSOLE] switch_time.c:1032 Calibrating timer, please wait... 2010-05-18 06:17:24.861063 [CONSOLE] switch_time.c:223 Test: 1000 Average: 2992 Step: 50 2010-05-18 06:17:24.861097 [WARNING] switch_time.c:227 Abnormally large timer gap 1992 detected! Do you have your kernel timer frequency set to lower than 1,000Hz? You may experience audio problems. any recommendations, other than drop to 5.3? which is not doable as the box is a 5k km away ... -- wasim h. baig | principal consultant | convergence pk | +92 30 0850 8070 | peace be upon you ... _______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org<mailto:FreeSWITCH-users@lists.freeswitch.org> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org -- Anthony Minessale II FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire AIM: anthm MSN:anthony_minessale@hotmail.com<mailto:MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<mailto:PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net<http://irc.freenode.net> #freeswitch FreeSWITCH Developer Conference sip:888@conference.freeswitch.org<mailto:sip%3A888@conference.freeswitch.org> googletalk:conf+888@conference.freeswitch.org<mailto:googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:+19193869900 !DSPAM:4bf2b5fb32931992811684!

Just so others reading this thread years down the line don't get the wrong idea, Tony's bark is much worse that his bite, he's really a big softie underneath, going out of his way, leaving no stone unturned to help folks out and ensure that any problem is identified quickly and accurately, and a fix provided pronto, proof being the yummy pudding FS is. Further comments inline below ... On Tue, May 18, 2010 at 20:35, Anthony Minessale < anthony.minessale@gmail.com> wrote:

Please heed as this is our official policy.

There is not an issue with CentOS 5.3. In fact CentOS 5.3 is the preferred platform (not 5.4 or 5.5 until we do proper testing and QA).

What you are describing is the result of poor hardware, (most likely the

motherboard)

If its the motherboard, then it should affect 5.3 as well. If not, then the issue is 5.4, and should (hopefully) be more addressable, rather than chucking out the hardware. For others with this issue, I'd recommend you drop down to 5.3 and verify the HW, if it goes away with 5.3, then lets see whats needed to fix it in 5.4/5.5. fwiw, i'm gonna try it on brand spanking new hpdl180g6 quad core xeon, will try both 5.3 and 5.4 FreeSWITCH is Free, your box is not, use the money you save not buying

FreeSWITCH to buy a decent box. Even a $500 box can work perfectly fine. Its not my intent to cater to inferior hardware when my goal is to push telephony into the future not get caught up in legacy drama with older boxes.

tony, these are the same boxes we did FS on way back when, remember the MGCP UA mode for * would keep crashing and you said screw it and started writing up FS instead, that was ages ago ... at least 5 years back ... so yeh, old boxes, dell poweredge 1450 ... point taken If you have altered the kernel to run at something besides 1000hz, change it

back to the default of 1000hz. If you get the error instantly it means the kernel is actually configured at the wrong freq. If the test runs for a long time and concludes it, it most likely means the box is not good enough to run FS.

sometimes it runs for long, sometimes its real fast ...

If you are not careful you may have installed the xen build of the kernel (check with uname -a) This can also cause timing issues.

stock kernel, 2.6.18-164.el5 If you see that timing gap warning (especially on centos if you have it set

to 1000hz) it means the motherboard cannot give the OS a good enough timing source to provide 1ms resolution which is mandatory and not very demanding.

I have been on dozens of perfectly happy boxes running CentOS 5.3 and I have been on a small handful of boxes that do what you describe and have seen that, once they are where they belong, (in the trash) the new inexpensive replacement does just fine.

:) ... 10-4 -wasim

On Tue, May 18, 2010 at 2:58 AM, Peter Olsson < peter.olsson@visionutveckling.se> wrote:

...

I also have this behaviour (timer gap around 2000). Upgraded yesterday to CentOS 5.5 (from 5.4), but the issue still remains. The system is just used for lab purposes, so I could easily fix SSH access if someone wanted to have a closer look.

I should also add – this is a 3-4 year old machine (workstation), but it seems this problem is related to newer CentOS-kernels, so I guess it shouldn’t be the issue here.

/Peter

*Från:* freeswitch-users-bounces@lists.freeswitch.org [mailto: freeswitch-users-bounces@lists.freeswitch.org] *För *Wasim Baig *Skickat:* den 18 maj 2010 06:20 *Till:* FreeSWITCH Users *Ämne:* [Freeswitch-users] centos 5.4 timer issue 1000hz, choppy voice at times

I know its been discussed previously on list, but has there been any progress to the timer issue on centos 5.4 32 bit

Linux pum2 2.6.18-164.el5 #1 SMP Thu Sep 3 03:33:56 EDT 2009 i686 i686 i386 GNU/Linux

[root@pum2 ~]# grep 1000 /boot/config-2.6.18-164.el5

CONFIG_HZ_1000=y

CONFIG_HZ=1000

i used a basic timer testing utility to confirm what we're actually getting

[root@pum2 ~]# ./timerchk

kernel timer interrupt frequency is approx. 1002 Hz

however, fs complains ...

2010-05-18 06:17:24.711360 [CONSOLE] switch_time.c:1032 Calibrating timer, please wait...

2010-05-18 06:17:24.861063 [CONSOLE] switch_time.c:223 Test: 1000 Average: 2992 Step: 50

2010-05-18 06:17:24.861097 [WARNING] switch_time.c:227 Abnormally large timer gap 1992 detected!

Do you have your kernel timer frequency set to lower than 1,000Hz? You may experience audio problems.

any recommendations, other than drop to 5.3? which is not doable as the box is a 5k km away ...

-- wasim h. baig | principal consultant | convergence pk | +92 30 0850 8070 | peace be upon you ...

!DSPAM:4bf2176432938985444358!

_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org

-- Anthony Minessale II

FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire

AIM: anthm MSN:anthony_minessale@hotmail.com <MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch

FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <sip%3A888@conference.freeswitch.org> googletalk:conf+888@conference.freeswitch.org<googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:+19193869900

_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org

-- wasim h. baig | principal consultant | convergence pk | +92 30 0850 8070 | peace be upon you ...

2-3 is fine. but the problem is when it's actually put to the test in userspace where you have many threads competing for cpu, that is where you have to measure the accuracy which is what that test during startup is doing. using FS to sleep 1000 microseconds and timing how close it actually is to that. voip can withstand jitter on most implementation as long as the running average of the packets is on the mark, the timing of each individual packet is less important, this worsens over latent connections. try this from the cli time_test 1000 you will see it come out + or - 100 easily (there is overhead for the code executing as well as other factors that can skew the results) then try it with a larger sample: time_test 1000 1000 time_test 1000 2000 etc the more times you try it the more likely the average in the long run will be accurate which is all we can hope for in our pseudo real time environment. most times if you try this with -nocal you will always get 2000 when you intended for 1000 which means all your 1ms sleeps are double length from what was requested. some newer kernels try to make it more accurate but by doing so they eat up so much cpu it's not worth it. For the most part semi-accurate 1ms timing is more than enough. On Tue, May 18, 2010 at 1:47 PM, Joseph L. Casale <jcasale@activenetwerx.com

wrote:

...

[root@pum2 ~]# ./timerchk kernel timer interrupt frequency is approx. 1002 Hz

So Anthony, What is the expected safe deviation fs would operate in where any audio issues I could assume would not be related to a shady clock.

On a variety of platforms I see 2-3 Hz variation but only on some does fs complain. Or could it be that this deviation alone isn't enough to quantify the quality of the clock?

Thanks for any input! jlc

_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org

-- Anthony Minessale II FreeSWITCH http://www.freeswitch.org/ ClueCon http://www.cluecon.com/ Twitter: http://twitter.com/FreeSWITCH_wire AIM: anthm MSN:anthony_minessale@hotmail.com <MSN%3Aanthony_minessale@hotmail.com> GTALK/JABBER/PAYPAL:anthony.minessale@gmail.com<PAYPAL%3Aanthony.minessale@gmail.com> IRC: irc.freenode.net #freeswitch FreeSWITCH Developer Conference sip:888@conference.freeswitch.org <sip%3A888@conference.freeswitch.org> googletalk:conf+888@conference.freeswitch.org<googletalk%3Aconf%2B888@conference.freeswitch.org> pstn:+19193869900