TOP发现系统负载整体很低,但CPU2的sys占用率很高在90%以上,查看当前正在运行的进程发现kipmi0进程占用率达到100%。
IPMI的初步认识:
IPMI是智能型平台管理接口(Intelligent Platform. Management Interface)的缩写,是管理基于Intel结构的企业系统中所使用的外围设备采用的一种工业标准,
该标准由英特尔/惠普/NEC/美国戴尔电脑和SuperMicro等公司制定。用户可以利用IPMI监视服务器的物理健康特征,如温度/电压/风扇工作状态、电源状态等。而且
更为重要的是IPMI是一个开放的免费标准,用户无需为使用该标准而支付额外的费用。
自1998 年,IPMI论坛创建了IPMI标准依赖,其已经得到了170 多家供应商的支持,使得其逐渐成为了一个完整地包括服务器和其他系统(如存储设备、网络和通信设备)的硬件管理规范,目前该标准最新版本为IPMI 2.0,该版本在原有基础上有了不少的改进,包括可以通过串口、Modem以及Lan等远程环境管理服务器系统(包括远程开关机),以及在安全、VLAN 和刀片支持等方面的提高。IPMI针对大量监 控、控制和自动回复服务器的作业,提供了智能型的管理方式。此标准适用于不同的服务器拓朴学,以及Windows、Linux、 Solaris、Mac或是混合型的操作系统。此外,由于IPMI可在不同的属性值下运作,即使服务器本身的运作不正常,或是由于任何原因而无法提供服 务,IPMI仍可正常运作。
查看IPMI相关的内核源码,发现09年提出的一个Patch(目前已合入主干,并做了再次修改):
-
[PATCH] limit CPU time spent in kipmid (version 4)
-
-
-
Signed-off-by: martin.wilck@xxxxxxxxxxxxxx
-
--- linux-2.6.29.4/drivers/char/ipmi/ipmi_si_intf.c 2009-05-19 01:52:34.000000000 +0200
-
+++ linux-2.6.29-rc8/drivers/char/ipmi/ipmi_si_intf.c 2009-06-04 15:30:34.855398091 +0200
-
@@ -297,6 +297,9 @@
-
static int force_kipmid[SI_MAX_PARMS];
-
static int num_force_kipmid;
-
-
+static unsigned int kipmid_max_busy_us[SI_MAX_PARMS];
-
+static int num_max_busy_us;
-
+
-
static int unload_when_empty = 1;
-
-
static int try_smi_init(struct smi_info *smi);
-
@@ -927,23 +930,56 @@
-
}
-
}
-
-
+#define ipmi_si_set_not_busy(timespec) \
-
+ do { (timespec)->tv_nsec = -1; } while (0)
-
+#define ipmi_si_is_busy(timespec) ((timespec)->tv_nsec != -1)
-
+
-
+static int ipmi_thread_busy_wait(enum si_sm_result smi_result,
-
+ const struct smi_info *smi_info,
-
+ struct timespec *busy_until)
-
+{
-
+ unsigned int max_busy_us = 0;
-
+
-
+ if (smi_info->intf_num < num_max_busy_us)
-
+ max_busy_us = kipmid_max_busy_us[smi_info->intf_num];
-
+ if (max_busy_us == 0 || smi_result != SI_SM_CALL_WITH_DELAY)
-
+ ipmi_si_set_not_busy(busy_until);
-
+ else if (!ipmi_si_is_busy(busy_until)) {
-
+ getnstimeofday(busy_until);
-
+ timespec_add_ns(busy_until, max_busy_us*NSEC_PER_USEC);
-
+ } else {
-
+ struct timespec now;
-
+ getnstimeofday(&now);
-
+ if (unlikely(timespec_compare(&now, busy_until) > 0)) {
-
+ ipmi_si_set_not_busy(busy_until);
-
+ return 0;
-
+ }
-
+ }
-
+ return 1;
-
+}
-
+
-
static int ipmi_thread(void *data)
-
{
-
struct smi_info *smi_info = data;
-
unsigned long flags;
-
enum si_sm_result smi_result;
-
+ struct timespec busy_until;
-
-
+ ipmi_si_set_not_busy(&busy_until);
-
set_user_nice(current, 19);
-
while (!kthread_should_stop()) {
-
+ int busy_wait;
-
spin_lock_irqsave(&(smi_info->si_lock), flags);
-
smi_result = smi_event_handler(smi_info, 0);
-
spin_unlock_irqrestore(&(smi_info->si_lock), flags);
-
+ busy_wait = ipmi_thread_busy_wait(smi_result, smi_info,
-
+ &busy_until);
-
if (smi_result == SI_SM_CALL_WITHOUT_DELAY)
-
; /* do nothing */
-
- else if (smi_result == SI_SM_CALL_WITH_DELAY)
-
+ else if (smi_result == SI_SM_CALL_WITH_DELAY && busy_wait)
-
schedule();
-
else
-
- schedule_timeout_interruptible(1);
-
+ schedule_timeout_interruptible(0);
-
}
-
return 0;
-
}
-
@@ -1213,6 +1249,11 @@
-
MODULE_PARM_DESC(unload_when_empty, "Unload the module if no interfaces are"
-
" specified or found, default is 1. Setting to 0"
-
" is useful for hot add of devices using hotmod.");
-
+module_param_array(kipmid_max_busy_us, uint, &num_max_busy_us, 0644);
-
+MODULE_PARM_DESC(kipmid_max_busy_us,
-
+ "Max time (in microseconds) to busy-wait for IPMI data before"
-
+ " sleeping. 0 (default) means to wait forever. Set to 100-500"
-
+ " if kipmid is using up a lot of CPU time.");
通过Patch的说明以及Patch中最后的参数介绍:当kipmid占用较多CPU时,可以将kipmid_max_busy_us设置100-500。
邮件原文如下,是为了降低kipmid的开销
-
Hi all,
-
-
I am sorry for the long silence. I am sending here a new version of my patch which takes into account Bela's suggestions (well, most of them).
-
I compiled and tested it with 2.6.29.4, the results are similar as before. By setting kipmid_max_busy_us to a value between 100 and 500, it is possible to bring down kipmid CPU load to practically 0 without loosing too much ipmi throughput performance.
-
-
Please give me some feedback whether this patch will get merged, and if not, what improvement is needed.
-
-
Regards
-
Martin
Kipmid的开销与其实现有关,暂时不深入究,不过可以通过Patch看到通过设置kipmid_max_busy_us来影响kipmid的调度策略,进而降低CPU的占用率。
看一下网上对ipmi占用CPU问题的说明:
Fix:不需要修复
No fix required. You should ignore increased CPU utilization as it has no impact on actual system performance.
利用空余的CPU资源进行一些接口自动调节的任务。
临时降低(立即生效,cpu占用率降到10%以内):
echo 100 > /sys/module/ipmi_si/parameters/kipmid_max_busy_us
永久性降低(修改配置文件,模块/系统重启生效)
To make the changes persistent you can configure the options for the ipmi_si kernel module.
Create a file in /etc/modprobe.d/, i.e./etc/modprobe.d/ipmi.conf, and add the following content:
# Prevent kipmi0 from consuming 100% CPU
echo "options ipmi_si kipmid_max_busy_us=100">/etc/modprobe.d/ipmi.conf
本文转自独弹古调 51CTO博客,原文链接:
http://blog.51cto.com/hunkz/1653806,如需转载请自行联系原作者