Question regarding mdadm.conf

Question regarding mdadm.conf

[Torsten E.]

Good morning! :)

End of january I had some problems with an faulty harddisk, which was 
part of an RAID1 (two mirroed disks containing 4 partitions).
I replaced it, and got some realy good instructions from Gordon Henderson.
Last week the same problem occured again, and I decided to replace the 
controller, too. Now its an PCI 4-port S-ATA controller, and everything 
runs fine till now.

My question now is:
the formerly used /etc/mdadm.conf had some lines like:
   ARRAY /dev/md1 level=raid1 num-devices=2
   UUID=66f80d1a:621442c5:2b6eabe8:3a9f5e54
    devices=/dev/sda6,/dev/sdb6
   ARRAY /dev/md0 level=raid1 num-devices=2
   UUID=06b9cdec:65448b34:3add2ff0:2180f5cd
    devices=/dev/sda5,/dev/sdb5

How does I get those UUID information, to add them to the new 
/etc/mdadm.conf?

Have a nice day
Torsten

Re: Question regarding mdadm.conf

[Lajber Zoltan]

Hi!

On Thu, 17 Feb 2005, Torsten E. wrote:

> How does I get those UUID information, to add them to the new
> /etc/mdadm.conf?

Try this one: mdadm --detail /dev/md1 | grep UUID

Bye,
-=Lajbi=----------------------------------------------------------------
 LAJBER Zoltan               Szent Istvan Egyetem,  Informatika Hivatal
   engineer: a mechanism for converting caffeine into designs.

Re: Question regarding mdadm.conf

[Michael Tokarev]

Lajber Zoltan wrote:
> Hi!
> 
> On Thu, 17 Feb 2005, Torsten E. wrote:
> 
> 
>>How does I get those UUID information, to add them to the new
>>/etc/mdadm.conf?
> 
> Try this one: mdadm --detail /dev/md1 | grep UUID

I'd say

   mdadm --detail --brief /dev/md1 | grep -v devices=

-- this will give you all information necessary for mdadm.conf,
you can just redirect output into that file.

Note the grep usage.  Someone will disagree with me here, but
there is a reason why to remove devices= line.  Without it,
output from mdadm looks like (on my system anyway):

ARRAY /dev/md1 level=raid1 num-devices=4 UUID=11e92e45:15fcc4a0:cf62e981:a79de494
    devices=/dev/sda1,/dev/sdb1,/dev/sdc1,/dev/sdd1

Ie, it lists all the devices which are parts of the array.
The problem with this is: if, for any reason (dead drive,
adding/removing drives/controllers etc) the devices will
change, and some /dev/sdXY will point to another device wich
is a part of some other raid array, mdadm will refuse to
assemble this array, saying something in a line of "the
UUIDs does not match, aborting".  Without the "devices="
part but with --scan option, mdadm will search all devices
by its own (based on the DEVICE line in mdadm.conf) - this
is somewhat slower as it will try to open each device in
turn, but safer, as it will find all the present components
no matter what.

Someone correct me if I'm wrong... ;)

/mjt

Re: Question regarding mdadm.conf

[Torsten E.]

Hi Michael,
Hi Lajber,

Thanks for your hints!

Michael Tokarev scribbled on 17.02.2005 08:14:
> Lajber Zoltan wrote:
> 
>> Hi!
>>
>> On Thu, 17 Feb 2005, Torsten E. wrote:
>>
>>> How does I get those UUID information, to add them to the new
>>> /etc/mdadm.conf?
>>
>> Try this one: mdadm --detail /dev/md1 | grep UUID
> 
> I'd say
> 
>   mdadm --detail --brief /dev/md1 | grep -v devices=

Didn't work for me ... but as I (hopefully!) understood the basic usage 
I run:
   mdadm --detail --brief /dev/md* >> /tmp/mdtest
After adding some more lines (DEVICE /dev/sda*, DEVICE /dev/sdb*, 
MAILADDR admin) I simply renamed it to /etc/mdadm.conf ... its not a 
good way, but for me its an useable (and so its good again ;)).

> -- this will give you all information necessary for mdadm.conf,
> you can just redirect output into that file.
> 
> Note the grep usage.  Someone will disagree with me here, but
> there is a reason why to remove devices= line.  Without it,
> output from mdadm looks like (on my system anyway):
> 
> ARRAY /dev/md1 level=raid1 num-devices=4 
> UUID=11e92e45:15fcc4a0:cf62e981:a79de494
>    devices=/dev/sda1,/dev/sdb1,/dev/sdc1,/dev/sdd1
> 
> Ie, it lists all the devices which are parts of the array.
> The problem with this is: if, for any reason (dead drive,
> adding/removing drives/controllers etc) the devices will
> change, and some /dev/sdXY will point to another device wich
> is a part of some other raid array, mdadm will refuse to
> assemble this array, saying something in a line of "the
> UUIDs does not match, aborting".  Without the "devices="
> part but with --scan option, mdadm will search all devices
> by its own (based on the DEVICE line in mdadm.conf) - this
> is somewhat slower as it will try to open each device in
> turn, but safer, as it will find all the present components
> no matter what.
> 
> Someone correct me if I'm wrong... ;)
> 
> /mjt

Have a nice day!! :)
Torsten

RE: Question regarding mdadm.conf

[Guy]

In the past, Neil has recommended using a device line like this:
DEVICE partitions

From "man mdadm.conf":
Alternatively, a device line can contain  the  word  partitions.
    This  will  cause mdadm to read /proc/partitions and include all
    devices and partitions found there-in.  mdadm does not  use  the
    names  from /proc/partitions but only the major and minor device
    numbers.  It scans /dev to find the name that matches  the  num-
    bers.

Guy

-----Original Message-----
From: linux-raid-owner@vger.kernel.org
[mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Torsten E.
Sent: Thursday, February 17, 2005 3:24 AM
To: linux-raid@vger.kernel.org
Subject: Re: Question regarding mdadm.conf

Hi Michael,
Hi Lajber,

Thanks for your hints!

Michael Tokarev scribbled on 17.02.2005 08:14:
> Lajber Zoltan wrote:
> 
>> Hi!
>>
>> On Thu, 17 Feb 2005, Torsten E. wrote:
>>
>>> How does I get those UUID information, to add them to the new
>>> /etc/mdadm.conf?
>>
>> Try this one: mdadm --detail /dev/md1 | grep UUID
> 
> I'd say
> 
>   mdadm --detail --brief /dev/md1 | grep -v devices=

Didn't work for me ... but as I (hopefully!) understood the basic usage 
I run:
   mdadm --detail --brief /dev/md* >> /tmp/mdtest
After adding some more lines (DEVICE /dev/sda*, DEVICE /dev/sdb*, 
MAILADDR admin) I simply renamed it to /etc/mdadm.conf ... its not a 
good way, but for me its an useable (and so its good again ;)).

> -- this will give you all information necessary for mdadm.conf,
> you can just redirect output into that file.
> 
> Note the grep usage.  Someone will disagree with me here, but
> there is a reason why to remove devices= line.  Without it,
> output from mdadm looks like (on my system anyway):
> 
> ARRAY /dev/md1 level=raid1 num-devices=4 
> UUID=11e92e45:15fcc4a0:cf62e981:a79de494
>    devices=/dev/sda1,/dev/sdb1,/dev/sdc1,/dev/sdd1
> 
> Ie, it lists all the devices which are parts of the array.
> The problem with this is: if, for any reason (dead drive,
> adding/removing drives/controllers etc) the devices will
> change, and some /dev/sdXY will point to another device wich
> is a part of some other raid array, mdadm will refuse to
> assemble this array, saying something in a line of "the
> UUIDs does not match, aborting".  Without the "devices="
> part but with --scan option, mdadm will search all devices
> by its own (based on the DEVICE line in mdadm.conf) - this
> is somewhat slower as it will try to open each device in
> turn, but safer, as it will find all the present components
> no matter what.
> 
> Someone correct me if I'm wrong... ;)
> 
> /mjt

Have a nice day!! :)
Torsten
-
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: Question regarding mdadm.conf

[GrantC]

On Thu, 17 Feb 2005 03:44:40 -0500, you wrote:

>In the past, Neil has recommended using a device line like this:
>DEVICE partitions
>
. . .
>>   mdadm --detail --brief /dev/md1 | grep -v devices=

peetoo:~$  mdadm --detail --brief /dev/md1
ARRAY /dev/md1 level=raid0 num-devices=2 UUID=732ae5f8:e5d69879:30435be8:aafb7e94
   devices=/dev/hda6,/dev/hdc6

peetoo:~$  mdadm --detail --brief /dev/md1 | grep -v " devices="
ARRAY /dev/md1 level=raid0 num-devices=2 UUID=732ae5f8:e5d69879:30435be8:aafb7e94

Is that what was intended?

Cheers
Grant.

Bad blocks

[Guy]

About 1 month ago the topic was bad blocks.  I have been monitoring the bad
blocks on my disks and I find I have had 3 new bad blocks since Jan 18.
Each on a different disk.  I have 17 disks, SEAGATE ST118202LC.

These bad blocks did not cause any problems with md.  I believe they were
readable or write errors, but re-mapped since I have AWRE and ARRE turned
on.

I don't know what is a normal rate, but based on the last month, I would
expect about 2.29 defects per disk per year.  I have had the disks in use
for about 2.5 years.  I have 81 defects.  That comes to 1.905 per disk per
year.  Not to far off the mark!  And I have replaced disks 2-3 times in 2.5
years.  The failed disks failed due to a cable problem, the disks may have
been just fine, but I took them apart for the magnets before I realized the
power cable was at fault.  The cable is now repaired.

Anyway, if these were read errors, md would have caused me lots of problems.
So, we need md to deal with bad blocks without kicking out the disk.

I have been using this command to monitor the disks:
sginfo -G /dev/sda | grep "in grown table"

My current bad block status (Defect list) for 17 disks:
0 entries (0 bytes) in grown table.
0 entries (0 bytes) in grown table.
8 entries (64 bytes) in grown table.
5 entries (40 bytes) in grown table.
0 entries (0 bytes) in grown table.
12 entries (96 bytes) in grown table.
20 entries (160 bytes) in grown table.
0 entries (0 bytes) in grown table.
6 entries (48 bytes) in grown table.
0 entries (0 bytes) in grown table.
6 entries (48 bytes) in grown table.
28 entries (224 bytes) in grown table.
0 entries (0 bytes) in grown table.
2 entries (16 bytes) in grown table.
3 entries (24 bytes) in grown table.
4 entries (32 bytes) in grown table.
0 entries (0 bytes) in grown table.

Does anyone know how many defects is considered too many?

Guy