Hello, i've been a long time user of Centos, mostly as a VmWare host OS.

I recently picked up 3x IBM Netvistas, 1.8Ghz P4, 512MB ram, 40GB IDE hard drive. I forget the exact model number, but I can provide that if needed.

I initially set up Centos 5.2 on these, updated them with yum, and installed Vmware Server on 2/3 of them. One just sat there idling doing nothing. After about 6-8 hours of uptime they would stop responding to SSH login requests. I would get a prompt for my username, after that it would sit and time out.

I could log in via keyboard at the physical machine, but I couldn't run any commands. Top would time out, sleep would hang, and pinging anything would return 1 packet/response and then it would hang.

I downgraded them to Centos 4.1, again 2 are running vmware server, one is a control that hasn't even been upgraded or touched. They all do the same thing, except I can log in via ssh after 10 hours of idling.

One peculiar behavior is demonstrated by running the 'time' command paired with sleep:

#######
[root@localhost ~]# hdparm -tT /dev/hda

/dev/hda:
Timing cached reads: 41360 MB in 3.71 seconds = 11161.98 MB/sec
Timing buffered disk reads: 3332 MB in 3.71 seconds = 899.22 MB/sec
[root@localhost ~]# time sleep 0.01

real 0m2.590s
user 0m0.000s
sys 0m0.000s
[root@localhost ~]# time sleep 0.05

real 0m0.359s
user 0m0.000s
sys 0m0.000s
[root@localhost ~]# time sleep 0.01

real 0m2.498s
user 0m0.000s
sys 0m0.000s
#######


In reality, time sleep 0.01 takes around 45-55 seconds to execute, timed using a physical clock. I am at a complete loss to explain what is going on here, or how to troubleshoot this. This only occurs on these IBM netvistas. I would also like to add, at one point these all used to run Windows Server 2003 without trouble. I have also performed a memory test using memtest86 on two of these machines. One tested clean with 1GB of ram (2x 512MB), another with 2x 256MB (512 total) had two memory errors. The other I haven't bothered with (the control machine)

Any ideas on what may be going on?

I downgraded them to Centos 4.1
???
4.9 is the current and only supported in the 4 series

With 512 meg ram i would expect them to be slow , VERY slow , and running a VM with 512 meg very very very slow .

With out knowing what was in the CentOS 5.6 logs that you erased when 4.1 was reinstalled ?
who knows

what did the logs say ?

0 members found this post helpful.

If you are implying that having 512MB of ram would cause sleep 0.01 to take a minute to complete (as I have described in the first post), that is complete nonsense. The host OS and Vmware Server use ~60MB of ram to boot to shell, and behave as expected for several hours before things begin to gum up. As I said in my first post, I am having trouble alone just getting the shell to respond in a period of time that does not have to be gauged by a sundial. My "control" machine that is un-updated, off the network, and without any other software installed than the base OS exhibits this same behavior.

If its absolutely necessary, I can provide the Centos 5 logs, but neither Centos 4 or 5 logs contained anything of alarm or interest, however recently I disabled ACPI via the grub bootloader, and now we have one valid error message to work with:


Jun 20 03:52:42 localhost kernel: Losing too many ticks!
Jun 20 03:52:42 localhost kernel: TSC cannot be used as a timesource.
Jun 20 03:52:42 localhost kernel: Possible reasons for this are:
Jun 20 03:52:42 localhost kernel: You're running with Speedstep,
Jun 20 03:52:42 localhost kernel: You don't have DMA enabled for your hard disk (see hdparm),
Jun 20 03:52:42 localhost kernel: Incorrect TSC synchronization on an SMP system (see dmesg).
Jun 20 03:52:42 localhost kernel: Falling back to a sane timesource now.


I've also attached the messages and dmesg logfile.

Attached Files

	messages.txt (57.9 KB, 11 views)
	dmesg.txt (12.2 KB, 12 views)

Any other ideas? :/