Hi guys I need help regarding Raid 0 becoming inactive on Raid 1+0 when one drive is removed.
We use centos 6.5 that is booting from the usb and the raid 1+0 is mounted as /data which holds virtualbox images and files.
We format the drives using gparted and created the raid using mdadm.
This is software-raid
so the setup is
/dev/sda and /dev/sdb = /dev/md1 (raid 1)
/dev/sdc and /dev/sdd = /dev/md2 (raid 1)
then we set them up as raid 0
/dev/md1 and /dev/md2 = /dev/md0 (raid 0)

I remember testing this a few years back that when /dev/sda is removed raid 0 is still active and just reports that a drive had failed.
same for sdb, sdc and sdd.
But recently when we tested it again we had found out that raid 0 is becoming inactive when any one of the four drives is removed.
So this is basically a change in behavior. I had tested this with the old kernel version on which we first tested the raid 1+0 and the new version via yum update and the result is the same.

Does anyone know why this happens?

Thanks!

After failing sda please post the outputs of:

Anyway, why do this instead of a single raid 10 array?

Hi that is what we are simulating. We are simulating a failed drive so we just turn off the server then pull one drive out then boot the system.
cat /proc/mdstat with all drives will produce something like
Personalities : [raid1] [raid0]
md125 : active raid0 md127[0] md126[1]
1953262592 blocks super 1.2 512k chunks

md126 : active raid1 sdc[2] sdd[1]
976631360 blocks super 1.2 [2/2] [UU]

md127 : active raid1 sdb[1] sda[2]
976631360 blocks super 1.2 [2/2] [UU]

But when we took one drive out it becomes

Personalities : [raid1] [raid0]
md125 : inactive raid0
1953262592 blocks super 1.2 512k chunks

md126 : active raid1 sdd[1]
976631360 blocks super 1.2 [2/2] [U_]

md127 : active raid1 sdb[1] sda[2]
976631360 blocks super 1.2 [2/2] [UU]

In theory raid 0 should still work as the raid 1 array only lost one drive. and therefore should still work.

We are now trying the single raid 10 array and also found out that it gives the same result.

Thanks!

If you feel you have sufficient documentation of your efforts, raise a bug against CentOS. I'm not a CentOS user, but you might not get a sympathetic response given the release you are running.