Time drifting problem on Dell x850
Robert von Bismarck
robert.vonbismarck at smart-telecom.ch
Wed Aug 9 04:27:13 CDT 2006
Hi,
First post on this list, so please bear with me if this is a known
issue.
We have several PE1850 and 2850, running a mix of SuSE 9.3 and CentOS
4.3.
All exhibit the same problem of having drifting clocks, by as much as 30
seconds in 24 hours.
We run ntpdate on all those systems once a day to a tier-2 time source
Here's an example of a 1850 running CentOS 4.3
Mar 6 12:00:14 log ntpdate[29745]: step time server aaa.bbb.ccc.ddd
offset 13.066364 sec
Mar 7 12:00:15 log ntpdate[30532]: step time server aaa.bbb.ccc.ddd
offset 13.074499 sec
Mar 8 12:00:15 log ntpdate[31437]: step time server aaa.bbb.ccc.ddd
offset 13.069962 sec
The same distribution with same patches on a similar HP box does not
have the problem.
A PE2650 with the same distrib is the worst of the lot with losses of up
to a minute every day. As this box is used to store and process
call-detail-records, timestamps must be precisely stored.
Things we have tried :
Recompile a vanilla kernel.org 2.6.17.6 kernel with timer set to either
100, 250 or 1000hz, which does not change a thing.
Upgrade the BIOS to A04 on the 1850, A05 on the 2850 and A20 on the
2650. Doesn't change anything either.
The only workaround we have found is to run ntpdate on the boxes every
hour, which is obviously not a *real* solution.
We're thankful for any further ideas.
Robert von Bismarck
Senior systems engineer
VTX SERVICES SA
A Smart-Telecom Group company
More information about the Linux-PowerEdge
mailing list