Time stopping/slowing in linux on TC4100

[Nick Gushlow]

I've found a problem happening on a load of linux servers I run.

Basically the clock is slowing right down so much that the clock changing
a second takes a minute or more. If I try to run top nothing happens, so I
assume the CPUs are being used insanely.

I forgot to check hwclock before rebooting, but as the clock is ok after
reboot (it was about 4 days out before reboot) so I assume it's ok.

Anybody have a clue as to what might be causing this? And where I could
look to find clues? I'm guessing it's a program not playing nice with the
hardware. I am steering away from hardware as the cause as I'm having this
problem on 7 linux boxes, yet no problems on 10 windows boxes; all
identical hardware.

For reference these boxes are HP TC4100 servers. Dual P3, SCSI, etc. I can
provide more hardware details if needed.

-- Nick

[P.T. Breuer]

In alt.os.linux Nick Gushlow <gushy79@hotmail.com> wrote: I've found a problem happening on a load of linux servers I run. Basically the clock is slowing right down so much that the clock changing a second takes a minute or more. If I try to run top nothing happens, so I assume the CPUs are being used insanely.

I assume so too, so look at the list of processes and SEE what is up.

Peter

[Nick Gushlow]

On Tue, 03 Aug 2004 14:58:12 +0200, P.T. Breuer wrote:
I assume so too, so look at the list of processes and SEE what is up.

I can't, neither top or ps respond when the systems are in this state.
Also I'm not able to soft reboot the box, I have to do a hard reset.

-- Nick

[Bill Unruh]

Nick Gushlow <gushy79@hotmail.com> writes:

]I've found a problem happening on a load of linux servers I run.

]Basically the clock is slowing right down so much that the clock changing
]a second takes a minute or more. If I try to run top nothing happens, so I
]assume the CPUs are being used insanely.

]I forgot to check hwclock before rebooting, but as the clock is ok after
]reboot (it was about 4 days out before reboot) so I assume it's ok.

]Anybody have a clue as to what might be causing this? And where I could
]look to find clues? I'm guessing it's a program not playing nice with the
]hardware. I am steering away from hardware as the cause as I'm having this
]problem on 7 linux boxes, yet no problems on 10 windows boxes; all
]identical hardware.

]For reference these boxes are HP TC4100 servers. Dual P3, SCSI, etc. I can
]provide more hardware details if needed.

run top and wait until it finally starts up so you can see what is going
on.
Or write a little shell script to run ps ax every minute or 30 sec and
putting the output into a /tmp file.
Then look to see what is hogging everything.
It sounds to me like it is swapping, which grinds everything to a halt. --
ie disk io is switching off all interrupts including the clock.

[P.T. Breuer]

In alt.os.linux Nick Gushlow <gushy79@hotmail.com> wrote: On Tue, 03 Aug 2004 14:58:12 +0200, P.T. Breuer wrote: I assume so too, so look at the list of processes and SEE what is up. I can't, neither top or ps respond when the systems are in this state.

Yes they do. (Hint: memory is reserved for root - keep a root
terminalopen).
Also I'm not able to soft reboot the box, I have to do a hard reset.

No you don't! Sysreq is always there.

Peter

[Nick Gushlow]

On Tue, 03 Aug 2004 17:00:57 +0000, P.T. Breuer wrote:
Yes they do. (Hint: memory is reserved for root - keep a root terminalopen).

Didn't know that. Although some of the boxes are remote so I'll have to
open up ssh access for root.
No you don't! Sysreq is always there.

Never used sysreq, just looked it up in the kernel docs. very handy! I can
use that on my local servers.


Thanks for your help. :-)

-- Nick

[Nick Gushlow]

On Tue, 03 Aug 2004 15:43:18 +0000, Bill Unruh wrote:
run top and wait until it finally starts up so you can see what is going on.

I've tended to run out of patience waiting for it! I'll hold out next time.
Or write a little shell script to run ps ax every minute or 30 sec and putting the output into a /tmp file.

That's a cracking idea, I'll do that.

Thanks,

-- Nick