OMSA and 2.4.18 kernal loses time

Dave Johnson dave at
Tue Dec 3 12:25:00 CST 2002


We too have seen drift, but we're not using OMSA.  
We also run 2.4.18-5 (and we have tried other 
kernels as well).  For us 2.4.18-5 seems to
also point to the environment, ie. does Dell
use a very low cost crystal to keep track of
system time which has more skew under higher
temperatures (which is sometimes proportionate
to how much load a fully packed machine is
under) ?

It would be interesting to see if this is 
a Linux kernel problem as we have swapped 
the really bad time "skewers" from production 
(1650s) and moved them to non-trading systems
because the SEC has regulations on what is 
acceptable machine time drift.

If your getting huge time skew, you may want to
run ntpd as opposed to cronjob ntpdates.
- Dave

Original Message:
OMSA and 2.4.18 kernal loses time 
Geoff French Geoff.French at 
Sun Dec 1 10:55:01 2002 

The use of OMSA and 2.4.18 kernels causes my newest poweredge systems
to lose timer ticks. They lose from 15 seconds to 1 hour a day.

I have six 1650's and four 2550's with Redhat 7.2 and OMSA 1.2
My one 2650 has Redhat 7.3 and OMSA 1.2 installed.

System time is kept accurate by using ntpdate twice a day to set the
system time. This normally keeps them within 10 seconds, the accuracy
required for my applications. I don't use ntpd. The hardware clocks keep
time just fine.

Since the upgrade of four of the 1650's to 2.4.18 kernels, thay all
lose almost 1 hour a day.

The two 1650's still using the last 2.4.9 kernels and OMSA 1.2 are fine.
They each gain about 4 seconds a day.

The 2550's (2 uniprocessor and 2 smp) using the 2.4.18 kernels
each gain about 15 seconds a day. Removed OMSA from two of the four
and there was no change in the 15 seconds gained each day.

I removed OMSA from the four 2.4.18 kernel 1650 systems, and they
are now keeping time approximately correctly (each gaining
about 13 seconds a day).

The 2650 was losing about 3 minutes a day. Removed OMSA and it is
now gains 17 seconds a day instead of losing 3 minutes.

It appears that there is some problem with 1650/2650 hardware and
OMSA 1.2 when using the 2.4.18 series kernels.

More information about the Linux-PowerEdge mailing list