[ntp:questions] Re: time reset, syncronisation lost and Large PPM values

mike michael.no.spam.cook at wanadoo.fr
Sun Oct 23 09:22:10 UTC 2005


While the smi interrupt could be the issue in your case, your original 
post indicates that the same problem are seen on other manafacturers 
servers (HP). HP may or may not use the same smi interrupt of course 
though I coudn't google any good fit.

However I have seen this on my own homebrew system and it turned out to 
be, as Tom said before, an institutionalised error had been
introduced. I had tried removing the drft file but still had the error.
Power off also didn't help.

The explanation is that the kernel also keeps a frequency drift value
that is persistent across boots. It is kept in /etc/adjtime. If this
file exists in your implementation, just removing it might be the 
answer. It fixed my problem.

The file is updated when the system is shutdown (at least on my 
implementation).

Mike


simbulu at gmail.com wrote:
> Thank you Tom for the guidance on how to restart ntpd in a proper way.
> I did that and will return my findings later (monday, say). Thanks for
> the suggestion of hardware problems - that got me gooing to new
> grounds.
> 
> I now have a feeling that this is a hardware problem related to SMI
> rather than APIC  (I always do whatever I can to remove any power
> management on any server), as reported by Pascal Blondé on Jun 4 2004:
> http://groups.google.com/group/comp.protocols.time.ntp/msg/00fe52bb908f6ed0
> 
> The saying (seemingly from Dell) is :
> "The PE2600 design inherently requires the BIOS generates an SMI
> (Systems Management Interrupt) ever 62ms. Since servers are not
> typically used for real-time applications,  customer impact should be
> low."
> 
> ntpd seems to be regarded as a real-time application, and real-time
> applications seems to be regarded as untypical on servers. In my world
> ntpd is mandatory on any server.
> 
> A google search for "poweredge 2600 ntpd smi" shows articles describing
> just what I see.  My boxes also use hyperthreading and SMP (2 x Xeon
> 2.8Ghz).
> 
> So now I would very much like to verify that the SMI thing really is
> the beast that kills my usage of ntpd. Does anyone have any idea of how
> this could be done ?
> 




More information about the questions mailing list