咋一个CPU使用100%,其他的还那么闲,怎么内核不会调度吗?
top - 10:10:24 up 35 min,2 users,load average: 6.63, 5.88, 4.53Tasks: 314 total, 6 running, 308 sleeping, 0 stopped, 0 zombie
Cpu0:2.0%us, 16.8%sy,0.0%ni, 80.9%id,0.0%wa,0.0%hi,0.3%si,0.0%st
Cpu1:1.0%us, 16.7%sy,0.0%ni, 81.9%id,0.0%wa,0.0%hi,0.3%si,0.0%st
Cpu2:2.4%us, 25.4%sy,0.0%ni, 71.9%id,0.0%wa,0.0%hi,0.3%si,0.0%st
Cpu3:0.0%us, 73.2%sy,0.0%ni,0.0%id,0.0%wa,0.0%hi, 26.8%si,0.0%st
Mem: 3334980k total,3202088k used, 132892k free, 12832k buffers
Swap: 0k total, 0k used, 0k free, 120920k cache
都是一些内核线程使用,软中断只发生在一个CPU,一个CPU用完了,其他CPU还很闲,怎么搞的? 这个可以调节网卡的中断配置来均衡网卡的中断。可以参考这个教程
http://www.mjmwired.net/kernel/Documentation/IRQ-affinity.txt
1 ChangeLog:
2 Started by Ingo Molnar <mingo@redhat.com>
3 Update by Max Krasnyansky <maxk@qualcomm.com>
4
5 SMP IRQ affinity
6
7 /proc/irq/IRQ#/smp_affinity and /proc/irq/IRQ#/smp_affinity_list specify
8 which target CPUs are permitted for a given IRQ source.It's a bitmask
9 (smp_affinity) or cpu list (smp_affinity_list) of allowed CPUs.It's not
10 allowed to turn off all CPUs, and if an IRQ controller does not support
11 IRQ affinity then the value will not change from the default of all cpus.
12
13 /proc/irq/default_smp_affinity specifies default affinity mask that applies
14 to all non-active IRQs. Once IRQ is allocated/activated its affinity bitmask
15 will be set to the default mask. It can then be changed as described above.
16 Default mask is 0xffffffff.
17
18 Here is an example of restricting IRQ44 (eth1) to CPU0-3 then restricting
19 it to CPU4-7 (this is an 8-CPU SMP box):
20
21 # cd /proc/irq/44
22 # cat smp_affinity
23 ffffffff
24
25 # echo 0f > smp_affinity
26 # cat smp_affinity
27 0000000f
28 # ping -f h
29 PING hell (195.4.7.3): 56 data bytes
30 ...
31 --- hell ping statistics ---
32 6029 packets transmitted, 6027 packets received, 0% packet loss
33 round-trip min/avg/max = 0.1/0.1/0.4 ms
34 # cat /proc/interrupts | grep 'CPU\|44:'
35 CPU0 CPU1 CPU2 CPU3 CPU4 CPU5 CPU6 CPU7
36 44: 1068 1785 1785 1783 0 0 0 0 IO-APIC-leveleth1
37
38 As can be seen from the line above IRQ44 was delivered only to the first four
39 processors (0-3).
40 Now lets restrict that IRQ to CPU(4-7).
41
42 # echo f0 > smp_affinity
43 # cat smp_affinity
44 000000f0
45 # ping -f h
46 PING hell (195.4.7.3): 56 data bytes
47 ..
48 --- hell ping statistics ---
49 2779 packets transmitted, 2777 packets received, 0% packet loss
50 round-trip min/avg/max = 0.1/0.5/585.4 ms
51 # cat /proc/interrupts |'CPU\|44:'
52 CPU0 CPU1 CPU2 CPU3 CPU4 CPU5 CPU6 CPU7
53 44: 1068 1785 1785 1783 1784 1069 1070 1069 IO-APIC-leveleth1
54
55 This time around IRQ44 was delivered only to the last four processors.
56 i.e counters for the CPU0-3 did not change.
57
58 Here is an example of limiting that same irq (44) to cpus 1024 to 1031:
59
60 # echo 1024-1031 > smp_affinity_list
61 # cat smp_affinity_list
62 1024-1031
63
64 Note that to do this with a bitmask would require 32 bitmasks of zero
65 to follow the pertinent one. 回复 2# crazyhadoop
看了这个没有问题,这个好像是调节硬件中断的吧?看几个CPU分配的挺平均的。不知道为什么软中断只在一个CPU发生?
页:
[1]