NTP unable to keep up with clock drift on 2600

Rainer Traut rainer.traut at epost.de
Thu Apr 15 13:48:00 CDT 2004


Hi,
how about:
echo "1" >/proc/irq/0/smp_affinity

I don't know when or where this was fixed but
this helps on some smp systems.
It prevents from the timer interrupt being handled by more than one cpu.

Rainer


Howie Frederickson wrote:
> Hi,
> 
> We're having clock drift problems on all of our 2600 boxes.  We're 
> trying to use NTP to keep the time on each of our systems in synch, but 
> on the 2600's, ntp keeps reporting that it is giving up.  We're not 
> having any such problems on our 1750's.
> Our setup currently has a 1750 as the local time server.  That server 
> uses several stratum 2 sources on the net to set its own clock, without 
> any issues.  Then, all local machines use that 1750 as the time server.  
> They are all on the same switch.   Our hope was that the LAN would be in 
> 'pretty good' synch with the rest of the world, but in 'very good' synch 
> with each other.  The 1750's are able to stay within ~10ms of each 
> other.   However, the 2600's can't seem to keep up.   One thing to note 
> is that the 2600's are our db servers (running Oracle 9i), and thus have 
> a heavier load than the 1750's.  But, the load is not _that_ heavy 
> (typically less than 1.5).  The /etc/ntp/drift file on the 2600's sets 
> itself to 500.00 (which I believe is the max)
> 
> I've tried updating the BIOS on the 2600's  (to A10), but that didn't help.
> 
> Anyone have any suggestions on how to deal with this?    Are there BIOS 
> settings, DMA settings, etc, that I should look at tweaking?
> 
> The 2600's are configured as follows:
> * dual 2.8 Xeon.
> * 4G ram
> * PERC4/Di  (romb)
> * PERC4/Dc ( PCI card)
> * 8x SCSI drives in 4 RAID1 volumes
> * Intel 2 port Gigabit nic (PCI card)
> 
> We're running RHAS 2.1, kernel 2.4.9-e.34enterprise, with NTP 
> ntp-4.1.2-1.AS21.1
> Hypethreading is disabled.
> 
> Here's a snippet from the /var/log/messages.   (The same message keep 
> coming every 30-60 mins)
> =================================
> Apr  6 23:03:23 starsky ntpd[12157]: synchronisation lost
> Apr  6 23:49:53 starsky ntpd[12157]: time reset 0.772319 s
> Apr  6 23:49:53 starsky ntpd[12157]: synchronisation lost
> Apr  7 00:29:46 starsky ntpd[12157]: time reset 0.747550 s
> Apr  7 00:29:46 starsky ntpd[12157]: synchronisation lost
> Apr  7 00:53:33 starsky ntpd[12157]: time reset 0.422760 s
> Apr  7 00:53:33 starsky ntpd[12157]: synchronisation lost
> Apr  7 01:41:01 starsky ntpd[12157]: time reset 0.727237 s
> Apr  7 01:41:01 starsky ntpd[12157]: synchronisation lost
> Apr  7 02:21:43 starsky ntpd[12157]: time reset 0.767331 s
> Apr  7 02:21:43 starsky ntpd[12157]: synchronisation lost
> Apr  7 02:43:26 starsky ntpd[12157]: time reset 0.302107 s
> Apr  7 02:43:26 starsky ntpd[12157]: synchronisation lost
> ==================================




More information about the Linux-PowerEdge mailing list