RAID 10 implementation

[TheGameAh]

Hi guys. Looking for some pointers here.

Trying to create a RAID10 file server, auxiliary storage really.

This is probably just a really stupid problem on my end, but can't figure it out for the life of me. Seams like basic RAID management really.

4 disks. All identically partitioned.

100 MB /boot, RAID1 with 2 spares (I hear you can't use RAID10 for /boot)
2 GB swap on each disk (not mirrored)
remainder of disk is /, RAID10

Everything installed fine.

So I remove a disk just to see what happens. Just testing everything really.

And the system can't boot. MDADM says it doesn't have enough drives to start the array. Well, obviously, I failed one on purpose. Shouldn't the system boot anyway? Isn't that the point of having a RAID, in case a drive fails?

[adamitj]

I'm not sure, but I think the clusters deployed with MDADM needs the original number of disks to work, ever. Only RAID 1 can work without the exactly number of disks AFAIK.

Try to start the RAID cluster and save a file. Remove one disk and format it. Then put it again on cluster and verify if synchronization is working on it.

[ryot]

Raid 10 should still work after a loss of a drive. You may try booting into the live CD and rebuilding it. After its rebuilt and in the OS, then remove a disk to test it.

[mam00th]

And the system can't boot. MDADM says it doesn't have enough drives to start the array. Well, obviously, I failed one on purpose. Shouldn't the system boot anyway? Isn't that the point of having a RAID, in case a drive fails?

AFAIK, raid1 wont function if one disk fail neither will raid 0 or 10. RAID5 and possibly raid 6 continue to run if one drive fail.

[ryot]

Here is a good overview of raid arrays.http://www.acnc.com/04_01_10.html Raid 10 has the same fault tolerance as raid 1, which is a mirror. The array will no longer be viable until rebuilt but the data is still there and can be booted as long as grub is installed on that drive.

[ryot]

And also, Raid 0 is the raid implementation that will not tolerate any failures. One drive dies.....all data gone!

[keld]

Raid10 will function in degraded mode if one disk fails.

See the howto on preventing against one failing disk using raid10 at http://linux-raid.osdl.org/index.php...a_failing_disk

[songshu]

your partitioning looks fine.
and RAID 10 should boot, i tested this on Debian Etch extensively detaching all the disks, not all at the the same time offcourse
http://www.songshu.org/index.php/setup-raid-10
this is what i did and this works.

i've also been hearing stories that Hardy RAID 1 has a bug, sounds like exactly the issue, you might just want to team up on launchpad for this

[windependence]

Just curious, why RAID 10? RAID 5 would work much better IMHO. You only need 3 disks to do RAID 5, and you can pull a drive with no problems although the array will be degraded.

-Tim

[songshu]

Just curious, why RAID 10? RAID 5 would work much better IMHO. You only need 3 disks to do RAID 5, and you can pull a drive with no problems although the array will be degraded.

-Tim

its a matter of dispute oftenly but with raid10 you can actually loose 2 disks (as long as long as its not the wrong 2) so slightly more secure.
at the cost of 50% diskspace instead of 30%

its faster, really it makes a difference (especially with linux raid in my experience)

most real problems with raid5 occure when you actually loose a disk and then need to hotswap, usually a disk that has been on stand by will suddenly be put under a big load and this is where percentage wise the moment things bad can happen to a raid5.

as far as i can tell raid10 will give you a little more performance and slightly more security at the cost of some diskspace and a little flexibility if you would like to increase the array at a later point.

this all providing your distubution of choice would ship without bugs in their mdadm package........