RAID1 /dev/md4 lost one disk partion /dev/sdb4 after reboot, WHY??

[Bruno Wolff III]

On Sun, Mar 29, 2009 at 10:00:07 +1100,
  L <yuanlux at gmail.com> wrote:
> Hi
> 
> I have RAID1 system,  the partition /dev/md4 is in raid1 with
> /dev/sda4 /dev/sdb4. In order to check /dev/sdb4 I remove it from
> /dev/md4

There are ways of checking disks without pulling them out of service.
smartctl can do surface scans while leaving the disk functional.

> The system did sync md4, this lasted ~ 6 hours. however, after reboot,
> /proc/mdstat shows /dev/md4 has one disk /dev/sda4 active.

6 hours is a long time for a resync. Was the system under heavy load?
Are an IO errors showing up in your logs?
 
> How to let system hold both partitions /dev/sda4 and /dev/sdb4 for
> /dev/md4 after reboot?

>From what you described it looked like you were correctly readding the
disk and waiting for the sync to complete before rebooting.