confusion about partitionable md arrays

confusion about partitionable md arrays

[Michael Schmitt]

Hi list,

since recent releases obviously it is possible to build an array and
partition that instead of building an array out of partitions. This was
somehow confusing but it worked in the first place. Now I moved the
array from one machine to another and... now it gets somehow strange. I
have nothing in /dev/md/ but /dev/md0. If I do fdisk -l it lists the
partitions /dev/md0p1 to /dev/md0p4, set to type 83 but there are no
devices under /dev/ or in /proc/partitions for them. I've read the
archives and googled around but there was no real solution just
different meanings on how it should be and how such things come. I'd
really appreciate a definite answer how that should work with
partitionable arrays and in best cases, what my problem may be here :)

At the end of this mail are the mdstat and mdadm outputs for reference

greetings and TIA
Michael

sinope:/home/mschmitt# mdadm --query /dev/md0
/dev/md0: 114.50GiB raid1 2 devices, 0 spares. Use mdadm --detail for
more detail.

sinope:/home/mschmitt# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid5] [raid4]
[raid6] [raid10]
md0 : active raid1 sda[0] sdb[1]
      120060800 blocks [2/2] [UU]

unused devices: <none>

sinope:/home/mschmitt# mdadm --examine /dev/sda
/dev/sda:
          Magic : a92b4efc
        Version : 00.90.00
           UUID : 4e832cd9:3e667457:cde5475c:8813896e
  Creation Time : Wed Oct  4 18:35:54 2006
     Raid Level : raid1
    Device Size : 120060800 (114.50 GiB 122.94 GB)
     Array Size : 120060800 (114.50 GiB 122.94 GB)
   Raid Devices : 2
  Total Devices : 2
Preferred Minor : 0

    Update Time : Tue Dec 26 01:43:01 2006
          State : clean
 Active Devices : 2
Working Devices : 2
 Failed Devices : 0
  Spare Devices : 0
       Checksum : 1dea11f0 - correct
         Events : 0.236


      Number   Major   Minor   RaidDevice State
this     0       8        0        0      active sync   /dev/sda

   0     0       8        0        0      active sync   /dev/sda
   1     1       8       16        1      active sync   /dev/sdb

Re: confusion about partitionable md arrays

[Bill Davidsen]

Michael Schmitt wrote:
> Hi list,
>
> since recent releases obviously it is possible to build an array and
> partition that instead of building an array out of partitions. This was
> somehow confusing but it worked in the first place. Now I moved the
> array from one machine to another and... now it gets somehow strange. I
> have nothing in /dev/md/ but /dev/md0. If I do fdisk -l it lists the
> partitions /dev/md0p1 to /dev/md0p4, set to type 83 but there are no
> devices under /dev/ or in /proc/partitions for them. I've read the
> archives and googled around but there was no real solution just
> different meanings on how it should be and how such things come. I'd
> really appreciate a definite answer how that should work with
> partitionable arrays and in best cases, what my problem may be here :)
>
> At the end of this mail are the mdstat and mdadm outputs for reference
>
>   
I'm not sure you have a problem, if this whole thing works correctly. 
However, there has been discussion about the implications of using whole 
drives instead of partitions to build your array. Having avoided that 
particular path I'm not going to rehash something I marginally 
understand, but some reading of post in the last few months may shed 
understanding.

-- 
bill davidsen <davidsen@tmr.com>
  CTO TMR Associates, Inc
  Doing interesting things with small computers since 1979

Re: confusion about partitionable md arrays

[Michael Schmitt]

Hi Bill,

this `thing´ works in another computer and as I moved it to another one,
I could not get it working. So the problem is... somewhere in between :)
As I had other non md-related problems in this computer (power supply
insufficient, system could not boot from hda anymore when sata disks
were connected to the pci-controller so I needed to unplug the sata
cables until grub has loaded, BIOS upgrade did not help), I moved the
array back in the original one, and it worked without a change.

But I noticed there a different behavior "Generating udev events for MD
arrays" showed up during boot, listing the partitions md0_d0p1 to
md0_d0p4 and of course /dev/md0_d0pN showed up. Then I cogitated about
all that a bit and came to some possible conclusions. Maybe the nameing
scheme for the arrays need to reflect somehow if it is a partibionable
array or not, if there are partitions and if udev needs to add the
appropriate devices... or not. Then I remembered some of the strange
things which happend when I did set up the array in the first place. I
tried to name the array "md0" but /dev/md_0" was generated too. /dev/md0
was somehow "dead" and my array was accessible as /dev/md0_d0. A bit
later as I tried to build an array with IDE discs in the "troublesome"
computer, I noticed similar effects. I tried to name the array
simple /dev/array1 but /dev/md127_d0 (iirc) was generated
too. /dev/array1 was dead and the raid was accessible at /dev/md127_d0.
So, whats all that? Even if I had read somehwere the name of the device
can be anything, it may be that that's not really true.

So the question remains the same, the confusion is persistent as the
superblocks in my arrays :) Any input is appreciated.

I already did read some mails in this list but did not get the point and
no real explanation to my question. I will read on, maybe I get
enlightenment.

greetings
Michael

Am Donnerstag, den 28.12.2006, 22:20 -0500 schrieb Bill Davidsen:
> Michael Schmitt wrote:
> > Hi list,
> >
> > since recent releases obviously it is possible to build an array and
> > partition that instead of building an array out of partitions. This was
> > somehow confusing but it worked in the first place. Now I moved the
> > array from one machine to another and... now it gets somehow strange. I
> > have nothing in /dev/md/ but /dev/md0. If I do fdisk -l it lists the
> > partitions /dev/md0p1 to /dev/md0p4, set to type 83 but there are no
> > devices under /dev/ or in /proc/partitions for them. I've read the
> > archives and googled around but there was no real solution just
> > different meanings on how it should be and how such things come. I'd
> > really appreciate a definite answer how that should work with
> > partitionable arrays and in best cases, what my problem may be here :)
> >
> > At the end of this mail are the mdstat and mdadm outputs for reference
> >
> >   
> I'm not sure you have a problem, if this whole thing works correctly. 
> However, there has been discussion about the implications of using whole 
> drives instead of partitions to build your array. Having avoided that 
> particular path I'm not going to rehash something I marginally 
> understand, but some reading of post in the last few months may shed 
> understanding.
> 

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: confusion about partitionable md arrays

[Bill Davidsen]

Michael Schmitt wrote:
> Hi Bill,
>
> this `thing´ works in another computer and as I moved it to another one,
> I could not get it working. So the problem is... somewhere in between :)
> As I had other non md-related problems in this computer (power supply
> insufficient, system could not boot from hda anymore when sata disks
> were connected to the pci-controller so I needed to unplug the sata
> cables until grub has loaded, BIOS upgrade did not help), I moved the
> array back in the original one, and it worked without a change.
>
> But I noticed there a different behavior "Generating udev events for MD
> arrays" showed up during boot, listing the partitions md0_d0p1 to
> md0_d0p4 and of course /dev/md0_d0pN showed up. Then I cogitated about
> all that a bit and came to some possible conclusions. Maybe the nameing
> scheme for the arrays need to reflect somehow if it is a partibionable
> array or not, if there are partitions and if udev needs to add the
> appropriate devices... or not. Then I remembered some of the strange
> things which happend when I did set up the array in the first place. I
> tried to name the array "md0" but /dev/md_0" was generated too. /dev/md0
> was somehow "dead" and my array was accessible as /dev/md0_d0. A bit
> later as I tried to build an array with IDE discs in the "troublesome"
> computer, I noticed similar effects. I tried to name the array
> simple /dev/array1 but /dev/md127_d0 (iirc) was generated
> too. /dev/array1 was dead and the raid was accessible at /dev/md127_d0.
> So, whats all that? Even if I had read somehwere the name of the device
> can be anything, it may be that that's not really true.
>
> So the question remains the same, the confusion is persistent as the
> superblocks in my arrays :) Any input is appreciated.
>
> I already did read some mails in this list but did not get the point and
> no real explanation to my question. I will read on, maybe I get
> enlightenment.
>   
I just reread the man page, and there is a paragraph on partitionable 
arrays following the "-a" option. Having warned you that I'm not 
experienced in this, I wonder if there is some interaction between 
device names created by "-amdp" and rules for udev, and how much of this 
partition info is carried in the superblock. There's a lot of discussion 
of names in that paragraph.

I also wonder if using the "-amdp" vs. just "-ap" will work differently. 
Hopefully Neil will have words of wisdom if we continue to thrash 
without a good understanding.
> greetings
> Michael
>
> Am Donnerstag, den 28.12.2006, 22:20 -0500 schrieb Bill Davidsen:
>   
>> Michael Schmitt wrote:
>>     
>>> Hi list,
>>>
>>> since recent releases obviously it is possible to build an array and
>>> partition that instead of building an array out of partitions. This was
>>> somehow confusing but it worked in the first place. Now I moved the
>>> array from one machine to another and... now it gets somehow strange. I
>>> have nothing in /dev/md/ but /dev/md0. If I do fdisk -l it lists the
>>> partitions /dev/md0p1 to /dev/md0p4, set to type 83 but there are no
>>> devices under /dev/ or in /proc/partitions for them. I've read the
>>> archives and googled around but there was no real solution just
>>> different meanings on how it should be and how such things come. I'd
>>> really appreciate a definite answer how that should work with
>>> partitionable arrays and in best cases, what my problem may be here :)
>>>
>>> At the end of this mail are the mdstat and mdadm outputs for reference
>>>
>>>   
>>>       
>> I'm not sure you have a problem, if this whole thing works correctly. 
>> However, there has been discussion about the implications of using whole 
>> drives instead of partitions to build your array. Having avoided that 
>> particular path I'm not going to rehash something I marginally 
>> understand, but some reading of post in the last few months may shed 
>> understanding.
>>
>>     

-- 
bill davidsen <davidsen@tmr.com>
  CTO TMR Associates, Inc
  Doing interesting things with small computers since 1979


-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html