Rtc: lost some interrupts at 1024Hz

Here is my problem. With the zaptel service started at some point in time something triggers an “rtc: lost some interrupts at 1024Hz.” message on the console. Now I understand ztdummy is used for timing on meetme conference rooms and IAX trunking. I’m not doing IAX trunking but I have begun using meetme. Strangely I can create and destroy numerous meetme rooms and not have a problem. Usually within several hours of continuos use the message begins to appear on the console. The message usually runs continuously, generating thousands of messages. The only way to end the constant stream of messages is to stop the zaptel service and then of course I lose the meetme functionality. Another side-effect of this is playback stops working. Callers now hear nothing but dead-air when calling into an IVR. This alone makes be a bit apprehensive about using ztdummy.

I have 7 Asterisk servers running on HP DL-360 G3 servers and every one has the same problem. I’ve compiled zaptel before asterisk. I’ve disabled ACPI. The G3 servers do not appear to have HPET option in BIOS - so no change there. I’ve tried older versions of zaptel with no luck. I have followed the advise of several similar postings in this forum as well as other articles posted in the voip-info wiki and have not made any forward progress with this problem. Any help would be greatly appreciated.

Here is /proc/driver/rtc

rtc_time : 16:28:20
rtc_date : 2008-12-18
rtc_epoch : 1900
alarm : **:00:00
DST_enable : no
BCD : yes
24hr : yes
square_wave : no
alarm_IRQ : no
update_IRQ : no
periodic_IRQ : yes
periodic_freq : 1024
batt_status : okay

Here is /proc/zaptel/1

Span 1: ZTDUMMY/1 “ZTDUMMY/1 (source: RTC) 1” (MASTER)

Upgraded to 1.4.23 and latest DAHDI and still having the same problem. I can’t imagine that I’m the only person on a HP Proliant DL-360 that is also trying to run Meetme. Does anyone know of a workaround, aside from spending $ on a Digium board just for clocking?

Well I’m a bit confused why your timing source is RTC in the first place, surely your server can support a high res timer. In ztdummy.c at the top there is a bit of logic to determine which timing source to use. The options are High-Res, High-Res w/ HPET, RTC, and Kernel (Linux26)

Try removing everything except the source you want, in this case “#define USE_HIGHRESTIMER”

See where that gets you…

what kernel do you have ?

Try doing a cat /proc/timer_list

If you don’t have it then your kernel does not support high_res timers.

I had to recompile Centos 5.2 with a vanilla kernel to get high_res timers to work as I only have digium b410p’s which use misdn.

I wasn’t running into the interrupt problem, but I had problems with playback app hanging for 5 minutes when RTC stopped sending ztdummy interrupts.