Fedora8 Dell Poweredge 2600 System Clock Jumps Forward

[mdebusschere]

Synposis:

Installed Fedora 8 from Live CD image onto a Dell Power Edge 2600 Server.
The disks are setup in raid 5 with a custom but typical load using boot
and LVM partition.

During boot it mentions it doesnt like clocksource, setting to HPET, but
otherwise no major problems booting. I have set the clocksource=hpet and
that appears to run without issue. Twice it has reported bad blocks
on root during boot, not sure why the system is usually shutdown clean
as far as I can tell..

Once the system is up, the system clock jumps ahead 4-5 minutes every so often.

Hence,
if I run date randomly a few seconds after each other I can see
the time jump 4 minutes, but if I run hwclock the clock runs as expected.

Of course during shutdown this fast moving system clock appears
to get set back to the hwclock.

I have tried every Kernel option I can think off, has been suggested etc, hence I
am asking here for suggestions I may have missed.

I have ntpd turned off.

[mdebusschere]

It appears that System time jumps forward every 14 seconds.
I read somewhere the kernel periodically checks & resets from
the hw clock at a similar interval ?

It doesn't really make sense as the hwclock keeps time normal
well at least over small period of time its not jumping minutes.

Anyone know how it might be adjusting or why during these
time checks ?

[mdebusschere]

Installed Fedora 7 as a test. It appears the clock is not
longer jumping forward. I assume this is kernel related.

[kilen]

Originally Posted by mdebusschere

Once the system is up, the system clock jumps ahead 4-5 minutes every so often.

I have the same problem but I'm running Fedora 7 on our PowerEdge 2600. I have just updated from Fedora Core 2. It may work fine for some time but occasinally the clock is steped several minits, and in /var/log/messages the line "Feb 13 15:06:00 adara kernel: Clocksource tsc unstable (delta = 238990016277 ns)" is printed!

When this occurs the server hangs and none of the services running on the PowerEdge is available on the network (dhcp, bind, httpd, svn etc). The wierd thing is that I only have to activate the server from screen saver state and its back on the network again.

This is a very big problem because the backup to DLT-tapes also failes when this happens.

Please, someone help! I don't wont to go back to Fedora Core 2.

[mdebusschere]

Try setting clocksource=hpet or as need as kernel param.

Also, I always though powersaving related items were causing the problems.
I don't recall if tried to disable all the possible scenarios.

It seems our system is stable with fedora 7. Sorry dont have a current
list of running daemons.

[kilen]

Originally Posted by mdebusschere

Try setting clocksource=hpet or as need as kernel param.

Ok, what kind of clocksource is that?

Also, I always though powersaving related items were causing the problems.
I don't recall if tried to disable all the possible scenarios.

Yesterday I rebooted the server with the kernel param acpi=off. And this morning the server is still running without problems. The time hasn't jumpt and backup finnished successfully. Hopefully the server will continue working now, but I'll have to wait a couple of days and see.

If not, I'll try the clocksource param.

[kilen]

My problem seems to be solved. After the server was booted with the kernel param "acpi=off", it has been running for severel days without any problems.

It seems like the acpi isn't compatible with the CPU.