All of lore.kernel.org
 help / color / mirror / Atom feed
* hard drives with "variable"  device names - mdadm raid assembly options setup
@ 2005-12-30 14:08 Mitchell Laks
  2005-12-31  0:44 ` Daniel Pittman
  0 siblings, 1 reply; 7+ messages in thread
From: Mitchell Laks @ 2005-12-30 14:08 UTC (permalink / raw)
  To: linux-raid

Hi,

I notice on my system that the hard drives "device names" can change.
Ie . 

The hard drives  are connected to the motherboard  SATA connectors or to PCI 
SATA cards. They get different "names" /dev/sda or /dev/sdc depending how 
many of the cards are "active". 

For instance the hard drives on the motherboard itself are called /dev/sde 
and /dev/sdf if there are hard drives on the 2 PCI cards and are 
called /dev/sda and /dev/sdb if there are none.

This is disconcerting, as I am setting up a remote site that I want to be able 
to visit rarely. I want to leave the "extra" hard drives on the PCI cards 
"off" for a few months, and am worried of FUD (fear uncertainty and doubt)
when I zoom in - to adjust the system - in a few months time when I come to 
connect the additional drives.

(Note: I am completely confused by udev which makes my raid life difficult - 
and I completely removed it so I am working with a static /dev/ 
directory :) )

I notice that there is a -u option based on the fixed uuid
mdadm -A /dev/md0 -u ?????

now I read in the man page:

If precisely one device is listed, but --scan is not given, then  mdadm
       acts  as  though --scan was given and identify information is extracted
       from the configuration file.

       The identity can be given with the --uuid  option,  with  the  --super-
       minor  option,  can be found  in the config file, or will be taken from
       the super block on the first component-device  listed  on  the  command
       line.

Now I am confused:  unfortunately this does not seem to help me: 
if it relys on the configuration file /etc/mdadm/mdadm.conf 
this config file - i create with the aid of mdadm --detail --scan 
seems to still have the (problematic) "variable" device names in it. 
I want only to use the uuid and not the "device name" which may change.
How can I deal effectively with my problem?  

I want to  set up my assemble line in my startup script so that all I need to 
do is 
1) connect the additional hard drives and 
2) uncomment a line in a startup script that will refer to the new hard drives 
only by their uuid so that I can be assured that the new device /dev/md1 will 
be the new drives and not a switcheroo? 

Thanks!

Mitchell

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: hard drives with "variable"  device names - mdadm raid assembly options setup
  2005-12-30 14:08 hard drives with "variable" device names - mdadm raid assembly options setup Mitchell Laks
@ 2005-12-31  0:44 ` Daniel Pittman
  2005-12-31  4:40   ` Max Waterman
                     ` (2 more replies)
  0 siblings, 3 replies; 7+ messages in thread
From: Daniel Pittman @ 2005-12-31  0:44 UTC (permalink / raw)
  To: linux-raid

Mitchell Laks <mlaks@verizon.net> writes:

G'day Mitchell.

> I notice on my system that the hard drives "device names" can change.
>
> The hard drives are connected to the motherboard SATA connectors or to
> PCI SATA cards. They get different "names" /dev/sda or /dev/sdc
> depending how many of the cards are "active".
>
> For instance the hard drives on the motherboard itself are called
> /dev/sde and /dev/sdf if there are hard drives on the 2 PCI cards and
> are called /dev/sda and /dev/sdb if there are none.

Well, one solution is udev, but since you prefer to avoid that...

[...]

> Now I am confused:  unfortunately this does not seem to help me: 
> if it relys on the configuration file /etc/mdadm/mdadm.conf 
> this config file - i create with the aid of mdadm --detail --scan 
> seems to still have the (problematic) "variable" device names in it. 

...you can simply remove the 'device names' variable, and list the
theoretically possible devices for MD stuff in the configuration:

DEVICE /dev/hd*
DEVICE /dev/sd*

That will include all the IDE and SCSI disks (including SATA), so they
will be scanned at an appropriate time.

Then, list your arrays:

ARRAY /dev/md2 level=raid1 num-devices=2 UUID=529d70fa:e5fe992b:ceb05593:bfcc6c25

That will cause mdadm to scan all those device entries (all the disks
and partitions) looking for an array with the right UUID, and assemble
it from all the components it finds.

Regards,
        Daniel


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: hard drives with "variable"  device names - mdadm raid assembly options setup
  2005-12-31  0:44 ` Daniel Pittman
@ 2005-12-31  4:40   ` Max Waterman
  2006-01-01 11:12     ` Daniel Pittman
  2005-12-31 12:01   ` Luca Berra
  2005-12-31 13:09   ` Gregory Seidman
  2 siblings, 1 reply; 7+ messages in thread
From: Max Waterman @ 2005-12-31  4:40 UTC (permalink / raw)
  To: linux-raid

Daniel Pittman wrote:
> Mitchell Laks <mlaks@verizon.net> writes:
> 
> G'day Mitchell.
> 
>> I notice on my system that the hard drives "device names" can change.
>>
>> The hard drives are connected to the motherboard SATA connectors or to
>> PCI SATA cards. They get different "names" /dev/sda or /dev/sdc
>> depending how many of the cards are "active".
>>
>> For instance the hard drives on the motherboard itself are called
>> /dev/sde and /dev/sdf if there are hard drives on the 2 PCI cards and
>> are called /dev/sda and /dev/sdb if there are none.
> 
> Well, one solution is udev, but since you prefer to avoid that...
> 
> [...]
> 
>> Now I am confused:  unfortunately this does not seem to help me: 
>> if it relys on the configuration file /etc/mdadm/mdadm.conf 
>> this config file - i create with the aid of mdadm --detail --scan 
>> seems to still have the (problematic) "variable" device names in it. 
> 
> ...you can simply remove the 'device names' variable, and list the
> theoretically possible devices for MD stuff in the configuration:
> 
> DEVICE /dev/hd*
> DEVICE /dev/sd*
> 
> That will include all the IDE and SCSI disks (including SATA), so they
> will be scanned at an appropriate time.
> 
> Then, list your arrays:
> 
> ARRAY /dev/md2 level=raid1 num-devices=2 UUID=529d70fa:e5fe992b:ceb05593:bfcc6c25
> 
> That will cause mdadm to scan all those device entries (all the disks
> and partitions) looking for an array with the right UUID, and assemble
> it from all the components it finds.

I am trying to do this with my 8 (currently only 7 since /dev/hdk is off 
line being replaced) disk raid5 array. mdadm --detail /dev/md0 gives :

"
/dev/md0:
         Version : 00.90.02
   Creation Time : Sat Aug  6 10:18:41 2005
      Raid Level : raid5
      Array Size : 976804480 (931.55 GiB 1000.25 GB)
     Device Size : 195360896 (186.31 GiB 200.05 GB)
    Raid Devices : 6
   Total Devices : 7
Preferred Minor : 0
     Persistence : Superblock is persistent

     Update Time : Sat Dec 31 12:35:32 2005
           State : clean
  Active Devices : 6
Working Devices : 7
  Failed Devices : 0
   Spare Devices : 1

          Layout : left-symmetric
      Chunk Size : 64K

            UUID : 15bfec75:595ac793:0914f8ee:862effd8
          Events : 0.838124

     Number   Major   Minor   RaidDevice State
        0      33        0        0      active sync   /dev/hde
        1      34        0        1      active sync   /dev/hdg
        2      56        0        2      active sync   /dev/hdi
        3       8        0        3      active sync   /dev/sda
        4       8       16        4      active sync   /dev/sdb
        5       8       32        5      active sync   /dev/sdc

        6       8       48        -      spare   /dev/sdd
"

I am confused why there are only 4 UUIDs, when there are 7 devices 
listed...what should I put in my mdadm.conf file, which is currently :

"
DEVICE /dev/hd[egik] /dev/sd[abcd]

ARRAY /dev/md0 level=raid5 num-devices=6 spares=1 
devices=/dev/hde,/dev/hdg,/dev/hdi,/dev/sda,/dev/sdb,/dev/sdc,/dev/sdd
"

What do I put for UUIDs for the 7 devices?

Max.


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: hard drives with "variable"  device names - mdadm raid assembly options setup
  2005-12-31  0:44 ` Daniel Pittman
  2005-12-31  4:40   ` Max Waterman
@ 2005-12-31 12:01   ` Luca Berra
  2005-12-31 13:09   ` Gregory Seidman
  2 siblings, 0 replies; 7+ messages in thread
From: Luca Berra @ 2005-12-31 12:01 UTC (permalink / raw)
  To: linux-raid

On Sat, Dec 31, 2005 at 11:44:32AM +1100, Daniel Pittman wrote:
>DEVICE /dev/hd*
>DEVICE /dev/sd*
>
i really find
DEVICE partitions
to be more useful than shell patterns.

L.

-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: hard drives with "variable"  device names - mdadm raid assembly options setup
  2005-12-31  0:44 ` Daniel Pittman
  2005-12-31  4:40   ` Max Waterman
  2005-12-31 12:01   ` Luca Berra
@ 2005-12-31 13:09   ` Gregory Seidman
  2 siblings, 0 replies; 7+ messages in thread
From: Gregory Seidman @ 2005-12-31 13:09 UTC (permalink / raw)
  To: linux-raid

On Sat, Dec 31, 2005 at 11:44:32AM +1100, Daniel Pittman wrote:
} Mitchell Laks <mlaks@verizon.net> writes:
} 
} G'day Mitchell.
} 
} > I notice on my system that the hard drives "device names" can change.
} >
} > The hard drives are connected to the motherboard SATA connectors or to
} > PCI SATA cards. They get different "names" /dev/sda or /dev/sdc
} > depending how many of the cards are "active".
} >
} > For instance the hard drives on the motherboard itself are called
} > /dev/sde and /dev/sdf if there are hard drives on the 2 PCI cards and
} > are called /dev/sda and /dev/sdb if there are none.
} 
} Well, one solution is udev, but since you prefer to avoid that...
[...]

I have an 8-tray SCSI hotswap rack, and I had a similar problem. I use
scsidev, from the scsitools package (under Debian, at least), to create
appropriately named block devices at boot before mdadm starts any md
devices. My /etc/scsi.alias:

id=0, hostname="Adaptec AIC7XXX", devtype=disk, alias=rack0
id=1, hostname="Adaptec AIC7XXX", devtype=disk, alias=rack1
id=2, hostname="Adaptec AIC7XXX", devtype=disk, alias=rack2
id=3, hostname="Adaptec AIC7XXX", devtype=disk, alias=rack3
id=8, hostname="Adaptec AIC7XXX", devtype=disk, alias=rack4
id=9, hostname="Adaptec AIC7XXX", devtype=disk, alias=rack5
id=10, hostname="Adaptec AIC7XXX", devtype=disk, alias=rack6
id=11, hostname="Adaptec AIC7XXX", devtype=disk, alias=rack7

Now, this is just matching by SCSI id (and SCSI host) since I don't want to
have to reconfigure on hotswap and the tray position controls the SCSI id,
but it can match by any combination of manufacturer, serial_number, model,
rev, wwid, lun, chan, hostname, hostid, and hostnum. Running scsidev -f
will create block devices in /dev/scsi; I have /dev/scsi/rack[0-7]. Then my
/etc/mdadm/mdadm.conf looks like this:

DEVICE /dev/scsi/rack[01234567]
ARRAY /dev/md0 devices=/dev/scsi/rack0,/dev/scsi/rack1,/dev/scsi/rack2,/dev/scsi/rack3,/dev/scsi/rack4,/dev/scsi/rack5,/dev/scsi/rack6,/dev/scsi/rack7

Very simple, very easy. The scsitools package includes scsiinfo, which
should work on SATA (or ieee1394, or USB) drives to retrieve serial
numbers. 

} Regards,
}         Daniel
--Greg


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: hard drives with "variable"  device names - mdadm raid assembly options setup
  2005-12-31  4:40   ` Max Waterman
@ 2006-01-01 11:12     ` Daniel Pittman
  2006-01-27 11:39       ` Matthias Urlichs
  0 siblings, 1 reply; 7+ messages in thread
From: Daniel Pittman @ 2006-01-01 11:12 UTC (permalink / raw)
  To: linux-raid

Max Waterman <davidmaxwaterman+gmane@fastmail.co.uk> writes:
> Daniel Pittman wrote:
>> Mitchell Laks <mlaks@verizon.net> writes:

[...]

>> Then, list your arrays:
>> ARRAY /dev/md2 level=raid1 num-devices=2
>> UUID=529d70fa:e5fe992b:ceb05593:bfcc6c25
>> That will cause mdadm to scan all those device entries (all the disks
>> and partitions) looking for an array with the right UUID, and assemble
>> it from all the components it finds.
>
> I am trying to do this with my 8 (currently only 7 since /dev/hdk is off
> line being replaced) disk raid5 array. mdadm --detail /dev/md0 gives :

[...]

>            UUID : 15bfec75:595ac793:0914f8ee:862effd8

[...]

> I am confused why there are only 4 UUIDs, when there are 7 devices
> listed...what should I put in my mdadm.conf file, which is currently 

You surely are confused, but there is only *one* UUID there, not four.

The UUID is presented as four blocks, with a ':' between them, by mdadm
for some reason beyond my understanding, but it is really just one
string.

Also, the UUID is per *MD* device, not per compontent (hda, sda) device.

So, you list the UUID for the array that mdadm told you as-is.

    Daniel


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: hard drives with "variable"  device names - mdadm raid assembly options setup
  2006-01-01 11:12     ` Daniel Pittman
@ 2006-01-27 11:39       ` Matthias Urlichs
  0 siblings, 0 replies; 7+ messages in thread
From: Matthias Urlichs @ 2006-01-27 11:39 UTC (permalink / raw)
  To: linux-raid

Hi, Daniel Pittman wrote:

> The UUID is presented as four blocks, with a ':' between them, by mdadm
> for some reason beyond my understanding

Human beings cannot remember 32 hex digits all at once, so if you ever
need to type the thing (or just compare them) it's helpful to have some
structure, even if arbitrary.

-- 
Matthias Urlichs   |   {M:U} IT Design @ m-u-it.de   |  smurf@smurf.noris.de
Disclaimer: The quote was selected randomly. Really. | http://smurf.noris.de
 - -
Duckies are fun!



^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2006-01-27 11:39 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2005-12-30 14:08 hard drives with "variable" device names - mdadm raid assembly options setup Mitchell Laks
2005-12-31  0:44 ` Daniel Pittman
2005-12-31  4:40   ` Max Waterman
2006-01-01 11:12     ` Daniel Pittman
2006-01-27 11:39       ` Matthias Urlichs
2005-12-31 12:01   ` Luca Berra
2005-12-31 13:09   ` Gregory Seidman

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.