Mdadm-faq

From Linux Raid Wiki
(Difference between revisions)
Jump to: navigation, search
(Another query...?)
 
(2 intermediate revisions by 2 users not shown)
Line 1: Line 1:
Debian has a [FAQ
+
== Debian mdadm FAQ ==
http://svn.debian.org/wsvn/pkg-mdadm/mdadm/trunk/debian/FAQ?op=file&rev=0&sc=0] on Linux raid. Some comments:
+
  
Answer 4b: The FAQ says that for a 4-disk raid10 array, you can survive two disk failures in half of the cases. The real figure is that in 2/3 of the cases you can survive.  
+
Debian has a [http://git.debian.org/?p=pkg-mdadm/mdadm.git;a=blob_plain;f=debian/FAQ;hb=HEAD FAQ] on Linux raid. Some comments:
  
 
Answer 6: The raid10 layout actually makes sense with only 2 disks, the FAQ says 3 disks are needed. Also "somewhat improved performance" is not adequeate, raid10,f2 offers something like twice the performance of normal RAID1+0 for sequential reading.
 
Answer 6: The raid10 layout actually makes sense with only 2 disks, the FAQ says 3 disks are needed. Also "somewhat improved performance" is not adequeate, raid10,f2 offers something like twice the performance of normal RAID1+0 for sequential reading.
Line 49: Line 48:
 
this one isn't so it must be a new device ?).
 
this one isn't so it must be a new device ?).
 
I should probably get it to re-use the slot number in that case.
 
I should probably get it to re-use the slot number in that case.
 +
 +
== What does the line ''Events : 0.123'' mean? ==
 +
Q:
 +
 +
When I do ''mdadm --query --detail /dev/md0'' there's this line that shows up somewhere in the listing: ''Events : 0.xyz'' . I've crawled the net for documentation about it, but I have no clue what the number represents.

Latest revision as of 15:06, 31 March 2009

[edit] Debian mdadm FAQ

Debian has a FAQ on Linux raid. Some comments:

Answer 6: The raid10 layout actually makes sense with only 2 disks, the FAQ says 3 disks are needed. Also "somewhat improved performance" is not adequeate, raid10,f2 offers something like twice the performance of normal RAID1+0 for sequential reading.

The FAQ references http://aput.net/~jheiss/raid10/ which actually does not refer to current Linux kernel raid10 technology, but to RAID1+0 versus RAID0+1 nested raid layout.

[edit] What does the output of mdadm --explain mean

Q:

On a raid1 with a 1.0 metadata I see:

  Array Slot : 3 (failed, failed, 0, 1)
  Array State : uU 2 failed

A:

Every device in a 1.x array is assigned a 'slot' number. This number is stable - it never changes.

Each device in the array also has a 'role' number indicating its current role in the array, which is either to be a spare or to have a position (0, 1, ...) in the array.

The output you produces says that this device occupies slot 3. It then notes that:

 the device which occupied slot 0 has failed
 the device which occupied slot 1 has failed
 the device which occupies slot 2 has role 0
 the device which occupies slot 3 has role 1

Then it shows you the state which indicated how the different roles are going.

  uU

means that both roles are 'up', and the 'this' device has the second role (capital U for 'this' device). Two devices have previously failed.

We should probably get rid of that '2 failed' bit, it isn't helpful. We should probably report 'empty' rather than 'failed' in the 'Array Slot' line.

Note that if you fail a devices, remove it, then add it back in such that it doesn't appear to be a re-add, it will be treated like a new device and get a new slot number. (after all the old device was faulty, this one isn't so it must be a new device ?). I should probably get it to re-use the slot number in that case.

[edit] What does the line Events : 0.123 mean?

Q:

When I do mdadm --query --detail /dev/md0 there's this line that shows up somewhere in the listing: Events : 0.xyz . I've crawled the net for documentation about it, but I have no clue what the number represents.

Personal tools