Hi All.
I have a server which uses RAID10 made of 4 partitions for / and boots from it. It looks like so:
mdadm -D /dev/md1 /dev/md1: Version : 00.90 Creation Time : Mon Apr 27 09:25:05 2009 Raid Level : raid10 Array Size : 973827968 (928.71 GiB 997.20 GB) Used Dev Size : 486913984 (464.36 GiB 498.60 GB) Raid Devices : 4 Total Devices : 4 Preferred Minor : 1 Persistence : Superblock is persistent
Update Time : Mon Apr 7 21:26:29 2014 State : clean Active Devices : 4 Working Devices : 4 Failed Devices : 0 Spare Devices : 0
Layout : near=2, far=1 Chunk Size : 64K
UUID : 1403e5aa:3152b3f8:086582aa:c95c4fc7 Events : 0.38695092
Number Major Minor RaidDevice State 0 8 6 0 active sync /dev/sda6 1 8 22 1 active sync /dev/sdb6 2 8 54 2 active sync /dev/sdd6 3 8 38 3 active sync /dev/sdc6
As far as I know raid10 is ~ "a raid0 built on top of two raid1" ( http://en.wikipedia.org/wiki/Nested_RAID_levels#RAID_1.2B0 - raid10). So I think that by default in my case:
/dev/sda6 and /dev/sdb6 form the first "raid1" /dev/sdd6 and /dev/sdc6 form the second "raid1"
So is it so that if I fail/remove for example: - /dev/sdb6 and /dev/sdc6 (different "raid1's") - the raid10 will be usable/data will be ok? - /dev/sda6 and /dev/sdb6 (the same "raid1") - the raid10 will be not usable/data will be lost?
I read in context of raid10 about replicas of data (2 by default) and the data layout (near/far/offset). I see in the output of mdadm -D the line "Layout : near=2, far=1" and am not sure which layout is exactly used and how it influences data layout/distribution in my case :|
I would really appreciate a definite answer which partitions I can remove and which I cannot remove at the same time because I need to perform some disk maintenance tasks on this raid10 array. Thanks for all help!
BR, Rafal.
On 4/7/2014 12:47 PM, Rafał Radecki wrote:
As far as I know raid10 is ~ "a raid0 built on top of two raid1" ( http://en.wikipedia.org/wiki/Nested_RAID_levels#RAID_1.2B0 - raid10). So I think that by default in my case:
/dev/sda6 and /dev/sdb6 form the first "raid1" /dev/sdd6 and /dev/sdc6 form the second "raid1"
So is it so that if I fail/remove for example:
- /dev/sdb6 and /dev/sdc6 (different "raid1's") - the raid10 will be
usable/data will be ok?
- /dev/sda6 and /dev/sdb6 (the same "raid1") - the raid10 will be not
usable/data will be lost?
I'm not sure you can verify that.... I would play it safe, and only drop one drive at a time.
On 2014-04-07, Rafa? Radecki radecki.rafal@gmail.com wrote:
I would really appreciate a definite answer which partitions I can remove and which I cannot remove at the same time because I need to perform some disk maintenance tasks on this raid10 array. Thanks for all help!
You're likely to get the most definitive answer from the linux RAID mailing list.
http://vger.kernel.org/vger-lists.html#linux-raid
Many of the md developers hang out there, and should know the correct answer. (I'm afraid I don't know it myself.)
--keith
On Tuesday, April 08, 2014 03:47 AM, Rafał Radecki wrote:
As far as I know raid10 is ~ "a raid0 built on top of two raid1" ( http://en.wikipedia.org/wiki/Nested_RAID_levels#RAID_1.2B0 - raid10). So I think that by default in my case:
No, Linux md raid10 is NOT a nested raid setup where you build a raid0 on top of two raid1 arrays.
/dev/sda6 and /dev/sdb6 form the first "raid1" /dev/sdd6 and /dev/sdc6 form the second "raid1"
So is it so that if I fail/remove for example:
- /dev/sdb6 and /dev/sdc6 (different "raid1's") - the raid10 will be
usable/data will be ok?
- /dev/sda6 and /dev/sdb6 (the same "raid1") - the raid10 will be not
usable/data will be lost?
The man page for md which has a section on "RAID10" describes the possibility of something is absolutely impossibe with a nested raid1+0 setup.
Excerpt: If, for example, an array is created with 5 devices and 2 replicas, then space equivalent to 2.5 of the devices will be available, and every block will be stored on two different devices.
So contrary to this statement: "RAID10 provides a combination of RAID1 and RAID0, and is sometimes known as RAID1+0.", linux md raid10 is NOT raid1+0. Is something entirely new and different but unfortunately called raid10 perhaps due to it being able to create a raid1+0 array and a different layout using similar concepts.
I read in context of raid10 about replicas of data (2 by default) and the data layout (near/far/offset). I see in the output of mdadm -D the line "Layout : near=2, far=1" and am not sure which layout is exactly used and how it influences data layout/distribution in my case :|
I would really appreciate a definite answer which partitions I can remove and which I cannot remove at the same time because I need to perform some disk maintenance tasks on this raid10 array. Thanks for all help!
If you want something that you can be sure about, do what I do. Make two raid1 md devices and then use them to make a raid0 device. raid10 is something cooked up by Neil Brown and but is not raid1+0. http://en.wikipedia.org/wiki/Linux_MD_RAID_10#LINUX-MD-RAID-10
The raid10 name is very misleading. I came to the same conclusion yesterday: for sake of clarity I will make two raid1 arrays and combine them into a raid0 ;)
Thanks for all info.
BR, Rafal.
2014-04-08 8:49 GMT+02:00 Christopher Chan <christopher.chan@bradbury.edu.hk
:
On Tuesday, April 08, 2014 03:47 AM, Rafał Radecki wrote:
As far as I know raid10 is ~ "a raid0 built on top of two raid1" ( http://en.wikipedia.org/wiki/Nested_RAID_levels#RAID_1.2B0 - raid10).
So I
think that by default in my case:
No, Linux md raid10 is NOT a nested raid setup where you build a raid0 on top of two raid1 arrays.
/dev/sda6 and /dev/sdb6 form the first "raid1" /dev/sdd6 and /dev/sdc6 form the second "raid1"
So is it so that if I fail/remove for example:
- /dev/sdb6 and /dev/sdc6 (different "raid1's") - the raid10 will be
usable/data will be ok?
- /dev/sda6 and /dev/sdb6 (the same "raid1") - the raid10 will be not
usable/data will be lost?
The man page for md which has a section on "RAID10" describes the possibility of something is absolutely impossibe with a nested raid1+0 setup.
Excerpt: If, for example, an array is created with 5 devices and 2 replicas, then space equivalent to 2.5 of the devices will be available, and every block will be stored on two different devices.
So contrary to this statement: "RAID10 provides a combination of RAID1 and RAID0, and is sometimes known as RAID1+0.", linux md raid10 is NOT raid1+0. Is something entirely new and different but unfortunately called raid10 perhaps due to it being able to create a raid1+0 array and a different layout using similar concepts.
I read in context of raid10 about replicas of data (2 by default) and the data layout (near/far/offset). I see in the output of mdadm -D the line "Layout : near=2, far=1" and am not sure which layout is exactly used and how it influences data layout/distribution in my case :|
I would really appreciate a definite answer which partitions I can remove and which I cannot remove at the same time because I need to perform some disk maintenance tasks on this raid10 array. Thanks for all help!
If you want something that you can be sure about, do what I do. Make two raid1 md devices and then use them to make a raid0 device. raid10 is something cooked up by Neil Brown and but is not raid1+0. http://en.wikipedia.org/wiki/Linux_MD_RAID_10#LINUX-MD-RAID-10 _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On 4/8/2014 12:35 AM, Rafał Radecki wrote:
The raid10 name is very misleading. I came to the same conclusion yesterday: for sake of clarity I will make two raid1 arrays and combine them into a raid0;)
Thanks for all info.
its striped mirrors, its just that it treats it all as one big raid rather than as distinct subraids.
having the separate mirrors and stripe can be an annoying complication, I"ve done it that way plenty of times before.
/dev/md2: /dev/md0: /dev/sda1 /dev/sdb1 /dev/md1: /dev/sdc1 /dev/sdd1
meh.
-
Christopher Chan -
John R Pierce -
Keith Keller -
Rafał Radecki