Home / server / Questions

Questions[interrupt](server)

Martin Hope
xddsg
Asked: 2012-01-13 06:42:57 +0800 CST
  • 2

Will the running the Windows dcdiag.exe command-line tool on a Server 2008 R2 system interfere with any services on that machine?

A small amount of CPU and network traffic usage is acceptable, but the system is live and applications on the machine must remain up and running without interruption.

domain-controller active-directory windows-server-2008-r2 interrupt
Martin Hope
Justin Holewinski
Asked: 2011-11-15 09:33:54 +0800 CST
  • 1

I'm wondering if anyone has seen this before, and could perhaps shed some light on a solution.

I'm running Red Hat Enterprise Linux 6.1, fully updated in a text-only server configuration, and the kacpid and kacpi_notify processes have started to take up ~70% and ~15% of CPU time, respectively. I've traced the problem to what looks like an insane amount of ACPI interrupts:

[root@centauri ~]# cat /proc/interrupts | grep acpi; sleep 5; cat /proc/interrupts | grep acpi
  9:     447753          0   32693472          0   IO-APIC-fasteoi   acpi
  9:     447753          0   32850749          0   IO-APIC-fasteoi   acpi

At over 30k interrupts a second, something is clearly wrong. Additionally, if I check another server with a very similar hardware configuration, the interrupt counts for this particular interrupt is always zero. Both systems are running Core i7 2600K processors with hyper-threading disabled. I've tried resetting all defaults in the BIOS, with no luck.

Any ideas what could cause such a high number of interrupts?

If I boot with acpi=off, this problem goes away. Besides power control, are there any drawbacks to running the system with ACPI disabled?

linux redhat acpi interrupt
Martin Hope
Linker3000
Asked: 2011-07-02 23:01:52 +0800 CST
  • 0

I have a Linux server running CentOS 5.6 (64-bit) that was running fine using the onboard NIC. For testing an HP PCI-e NIC was installed too but the server wouldn't respond on the network through eth0 or eth1 and the server then threw a debug message that it was disabling INT 50 so I presumed there was some conflict between the two NICs.

Unfortunately, the server will now not communicate through either the onboard NIC (HP card removed) or the HP card with the onboard NIC disabled - in both cases the server confirms it has an eth0 card and ifconfig returns sensible values, but the server's not PINGable and you can't PING anything on the network. If the NIC is set to use DHCP it won't get an address. The patch lead has been tested and also replaced.

I get the feeling that this is still an interrupt-related issue (lspci shows the NIC on INT 50), so when Linux/CentOS disables an interrupt is this written down in a config file somewhere that now needs to be manually edited?

Any thoughts appreciated before I try resetting the BIOS config back to factory defaults and/or begin to suspect that something's been damaged.

linux centos nic interrupt