Felix's Second Life 電腦數碼世界

Vmware ESX 4上虛擬機 Redhat 5.2（CentOS 5.2）啟動在Starting udev 停幾個小時
最近在vmware ESX上安裝的Redhat 5.2或者CentOS 5.2基本上都碰到了這個問題
安裝完全完成後啟動是在Starting udev會需要幾個小時才能過去。
網上找到很多說法比方有說讓改"/etc/udev/rules.d/99-vmware-scsi-udev.rules" 文件的，其實這並不是hang住的原因，真正的元兇是：
Clocksource is the name of the new timekeeping subsystem introduced in the Linux 2.6.18 kernel. In this subsystem there are several clocksources that can be used to read the time. One of them is the tsc clocksource which uses the Time Stamp Counter that is present on each processor and can be used to track how much time has elapsed. The TSCs on the different processors are not necessarily perfectly in sync with each other, so time can appear to go backward if the TSC is read on one processor and then another processor. Early versions of the tsc clocksource did not handle this case, which can lead to the guest operating system not responding. Due to the different timing behavior of running in a virtual machine, this can be observed more frequently in a virtual machine.
經驗證確實如此：如果是單個vCPU,則Starting udev很快就能過去
具體解決方法：
當starting udev停住很久過不去的時候就(1)把虛擬機的電源關閉，(2)然後編輯配置,改為單個vCPU
(3)啟動系統 （4）編輯/etc/grub.conf在kernel一行最後加上clocksource=acpi_pm
這個在vmware官方有說明：
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1007020
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1006427