[asterisk-users] hardware clock drift and CDR

[Gordon Henderson]

On Mon, 26 Apr 2010, Vieri wrote:

> I ran the following and it supposedly updated my system time while ntpd was running:
>
> # ps ax | fgrep ntp
> 1256 ?        Ss     0:00 /usr/sbin/ntpd -p /var/run/ntpd.pid -u ntp:ntp
> 1623 pts/14   S+     0:00 fgrep ntp
>
> # ntpdate -b -u pool.ntp.org
> 26 Apr 19:41:18 ntpdate[2791]: step time server 163.117.131.239 offset 0.142263 sec

Steves posted the reason - the -u flag causes it to bypass the normal 
ports, and so does work in this instance.

> By the way, as a side question, on another server I see this:
>
> # ntpq -c peers
>     remote           refid      st t when poll reach   delay   offset  jitter
> ==============================================================================
> inf-srv1.hospit .LOCL.           1 u   56   64  377    0.314  21755.8   7.634
>
> Not sure what LOCL means but I'll refer to the NTP docs (inf-srv1 is my 
> LAN Windoze time server).

It may mean that it's using it's internal clock as the master source. If 
so, then it's trust it as far as I could spit a rat...

Try this:

   ntpq
   host inf-srv1

(or it's IP addresS)

   peers

and find out what peers it's using.

It's just possible that your server is actually more accurate that your 
LAN server... Give your server a few more peers and find out - just list 
pool.ntp.org in the /etc/ntp.conf file a few times (and restart ntpd)

> Anyway, back to the faulty new server (which reports a stratum of 3 
> after ntpd has been running for a while and sync'ing to pool.ntp.org):

The stratus is just how far it is away from stratum 1 - which is deemed to 
be synchronised to "true" time - usually derived from GPS, local atomic 
clock or MSF type radio. (I used to run an MSF clock synced NTP server for 
a while) So a host synchronised to a stratum 1 server will be at stratum 
2, and hosts synchronised to a stratum 2 server will be at stratum 3. If 
you synchronise to a mixture, then your host will be somewhere in the 
range, depending on how good it reckons the other are...

> it's supposed to be a good motherboard (Asus) but I'm running a 
> relatively "old" kernel (2.6.23). Googling around suggests me to try to 
> boot with "noapic" if I keep seeing my clock drift so much.
>
> # more /proc/interrupts
>           CPU0       CPU1       CPU2       CPU3
>  0:        103          0          0          1   IO-APIC-edge      timer
>  1:       2151          0          0          9   IO-APIC-edge      i8042
>  4:   12772543   13217932    9603064    7661766   IO-APIC-edge      serial

That's a rather high number of serial interrupts... Do you have a serial 
console, or using the serial link with Linux HA?

In-general, I like ASUS motherboards though and use them a lot myself.

>  8:          1          0          1          0   IO-APIC-edge      rtc
>  9:          0          0          0          1   IO-APIC-fasteoi   acpi
> 12:          0          0          0          4   IO-APIC-edge      i8042
> 14:       2234      73664          0       2470   IO-APIC-edge      ide0
> 16:   28322780   51914617   40744985   39615361   IO-APIC-fasteoi   eth0
> 17:   63242610   42157366   43790794   48255583   IO-APIC-fasteoi   eth1
> 18:    1348544          0          0          1   IO-APIC-fasteoi   eth2
> 20:    9006839    8244295    6076595    4923525   IO-APIC-fasteoi   ahci
> 21:  162750903  140985080  176469550  166839225   IO-APIC-fasteoi   wcte12xp0
> 22:   16662710   18210608   12053147   12739782   IO-APIC-fasteoi   HFC-multi
> NMI:          0          0          0          0
> LOC:   64546905   64546897   64546897   64546897
> ERR:          0
> MIS:          0
>
> I have 3 PCI cards: 1 PRI, 1 quad BRI, 1 dual ethernet.
>
> Could booting with "noapic" help?

Doubt it, but iy's worth a try. Personally, I'd try more NTP hosts first. 
(Especially knowing you're syncing to a windoze host ;-)

> What about my PCI devices? Will they be stable even with "noapic"?

> The reason I got this new mobo is that the previous hardware froze the 
> system with a kernel crash. In fact, I rsync'ed to this new hardware (so 
> identical system software) and it has been running flawlessly for more 
> than a week now, while it used to crash/freeze once a day (another Asus 
> board, by the way). My only problem now is with the d@!mned clock...
>
> As far as syslog messages, I don't see anything wrong. No errors whatsoever.
>
> Thanks for your time. I'll try to boot with noapic and cross my fingers.

Good luck..

What may also help is compiling a custom kernel for your hardware - it's 
what I do by default, but I appreciate that's not for everyone, however it 
is the best way to make sure you have the kernel tuned exactly to your 
hardware needs.

Gordon