asked on

RHEL 5.6 and Interrrupts

I am trying to investigate if a machine is hitting its limit of interrupts.

I get the following :

[root@WEST5 ~]# ps -ef | grep ksoftirqd
root         3     1  0 May02 ?        00:00:00 [ksoftirqd/0]
root         6     1  0 May02 ?        00:00:00 [ksoftirqd/1]
root         9     1  0 May02 ?        00:00:00 [ksoftirqd/2]
root        12     1  0 May02 ?        00:00:00 [ksoftirqd/3]
root        15     1  0 May02 ?        00:00:00 [ksoftirqd/4]
root        18     1  0 May02 ?        00:00:00 [ksoftirqd/5]
root        21     1  0 May02 ?        00:00:00 [ksoftirqd/6]
root        24     1  0 May02 ?        00:00:00 [ksoftirqd/7]
root        27     1  0 May02 ?        00:00:00 [ksoftirqd/8]
root        30     1  0 May02 ?        00:00:00 [ksoftirqd/9]
root        33     1  0 May02 ?        00:00:00 [ksoftirqd/10]
root        36     1  0 May02 ?        00:00:00 [ksoftirqd/11]
root        39     1  0 May02 ?        00:00:00 [ksoftirqd/12]
root        42     1  0 May02 ?        00:00:00 [ksoftirqd/13]
root        45     1  0 May02 ?        00:00:00 [ksoftirqd/14]
root        48     1  0 May02 ?        00:00:00 [ksoftirqd/15]
root        51     1  0 May02 ?        00:00:00 [ksoftirqd/16]
root        54     1  0 May02 ?        00:00:00 [ksoftirqd/17]
root        57     1  0 May02 ?        00:00:00 [ksoftirqd/18]
root        60     1  0 May02 ?        00:00:00 [ksoftirqd/19]
root        63     1  0 May02 ?        00:00:00 [ksoftirqd/20]
root        66     1  0 May02 ?        00:00:00 [ksoftirqd/21]
root        69     1  0 May02 ?        00:00:00 [ksoftirqd/22]
root        72     1  0 May02 ?        00:00:00 [ksoftirqd/23]
root     12778 12715  0 11:21 pts/4    00:00:00 grep ksoftirqd

Open in new window

Does this indicate a normal state (I am assming it does)

What would it look like if I had reached a state where I had too many interrupts which were slowing the machine down ?

dead_philosopher

Interrupt storms are perceived as general system slowness with increased load. They differ from thrashing or out of memory style slowness in no utilization of swap and disk I/O. The perception of impact will really depend on the system use (desktop versus server), both will essentially be slow and burst to normal but based on applications, etc. your problem statements can vary.

Remediation:
cat /proc/interrupts to identify potential candidates for pinning tasks (taskset) and consider using isolcpus (isolcpus=4,5) in your system grub.conf with taskset to isolate CPUs for designated tasks.

ASKER CERTIFIED SOLUTION

lnkevin

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial