Anaconda Incorrectly Detects dmraid

[Christopher Boumenot]

Greetings,

We are seeing an issue with anaconda-11.1.2.36 (CentOS 5.0), where it 
incorrectly detects dmraid on our Dell 2900s (and, 1950s, 2950s).  (We 
were able to track it down to that issue based on the message "Specified 
nonexistent disk sda in partition command," and /tmp/anaconda.log.)  I 
am not sure why Anaconda detects dmraid.  I suspect it looks for a piece 
of metadata somewhere on the disk, and by pure hapstance our data 
matches the signature of a dmraid disk set.  Has anyone seen this 
before?  Any ideas on how to fix it?  We have not seen this issue with 
RHEL 4u2, or CentOS 4u4, and we have been running/imaging those for the 
past 1.5 years.

Here's a example log from two identical machines, one with the issue, 
and one without.

-- Falsely detects dmraid. --

DEBUG   : starting dmraids
DEBUG   : self.driveList(): ['sda', 'sdb']
DEBUG   : DiskSet.skippedDisks: []
DEBUG   : DiskSet.skippedDisks: []
DEBUG   : starting all dmraids on drives ['sda', 'sdb']
DEBUG   : scanning for dmraid on drives ['sda', 'sdb']
DEBUG   : got raidset <block.device.RaidSet instance at 0xb7d173ac> (sda 
sdb)
DEBUG   :   valid: True found_devs: 2 total_devs: 2
DEBUG   : adding 
mapper/ddf1_44656c6c202020201028001510281f0332b494acaae97d7b to isys cache
DEBUG   : adding 
mapper/ddf1_44656c6c202020201028001510281f0332b494acaae97d7b to isys cache()
DEBUG   : adding sda to dm cache
DEBUG   : removing sda from isys cache
DEBUG   : adding sdb to dm cache
DEBUG   : removing sdb from isys cache

## it detected dmraid?!?!
-->>DEBUG   : starting raid <block.device.RaidSet instance at 
0xb7d173ac> with mknod=True

DEBUG   : done starting dmraids.  Drivelist: 
['mapper/ddf1_44656c6c202020201028001510281f0332b494acaae97d7b']
INFO    : Running kickstart %%pre script(s)
INFO    : All kickstart %%pre script(s) have been run

-- Correctly, does not detect dmraid. --

DEBUG   : starting dmraids
DEBUG   : self.driveList(): ['sda', 'sdb']
DEBUG   : DiskSet.skippedDisks: []
DEBUG   : DiskSet.skippedDisks: []
DEBUG   : starting all dmraids on drives ['sda', 'sdb']
DEBUG   : scanning for dmraid on drives ['sda', 'sdb']
DEBUG   : done starting dmraids.  Drivelist: ['sda', 'sdb']
INFO    : Running kickstart %%pre script(s)
INFO    : All kickstart %%pre script(s) have been run

In the meantime, to get around this issue I just commented out all calls 
to startDmRaid().  I will continue to decipher the code (anaconda, 
python-pyblock, dmraid-devel) , but I wanted to ask here first.

Thanks,
Christopher

[HARA Hiroshi]

Hi,

I also encountered such kind of issue on ata(ICH6/7R) sometimes
and I also think there are something wrong in metadata of disk.

to find why this kind of problem happen,
we need to investigate libs of dmraid and python-pyblock.

so if you just do not want to use dmraid, erase metadata from disk
or just use "nodmraid" option. (boot:linux nodmraid)

In my case I erased metadata with following way.
--------------------------
# dmraid -E -r /dev/sda
--> In my case, this command can not work well, so I did the following
--------------------------

--------------------------
# dmraid -r /dev/sda
/dev/sda: ddf1, ".ddf1_disks", GROUP, ok, 976496964 sectors, data@ 0

# dmraid -D -r /dev/sda
/dev/sda: ddf1, ".ddf1_disks", GROUP, ok, 976496964 sectors, data@ 0

# ls -l sda_ddf1.*
-rw------- 1 root root 6144 5月 10 04:06 sda_ddf1.dat
-rw------- 1 root root 13 5月 10 04:06 sda_ddf1.offset
-rw------- 1 root root 10 5月 10 04:06 sda_ddf1.size

# dd if=/dev/zero of=/dev/sda seek=`cat sda_ddf1.offset` bs=1 count=6144
6144+0 records in
6144+0 records out
6144 bytes (6.1 kB) copied, 0.0217024 seconds, 283 kB/s

# dmraid -r /dev/sda
No RAID disks and with names: "/dev/sda"
--------------------------

If anybody have any information about dmraid, would you inform us,please?
actually, I'm not sure the stability of dmraid...

thank you,

Christopher Boumenot wrote:
> Greetings,
>
> We are seeing an issue with anaconda-11.1.2.36 (CentOS 5.0), where it 
> incorrectly detects dmraid on our Dell 2900s (and, 1950s, 2950s).  (We 
> were able to track it down to that issue based on the message "Specified 
> nonexistent disk sda in partition command," and /tmp/anaconda.log.)  I 
> am not sure why Anaconda detects dmraid.  I suspect it looks for a piece 
> of metadata somewhere on the disk, and by pure hapstance our data 
> matches the signature of a dmraid disk set.  Has anyone seen this 
> before?  Any ideas on how to fix it?  We have not seen this issue with 
> RHEL 4u2, or CentOS 4u4, and we have been running/imaging those for the 
> past 1.5 years.
>
> Here's a example log from two identical machines, one with the issue, 
> and one without.
>
> -- Falsely detects dmraid. --
>
> DEBUG   : starting dmraids
> DEBUG   : self.driveList(): ['sda', 'sdb']
> DEBUG   : DiskSet.skippedDisks: []
> DEBUG   : DiskSet.skippedDisks: []
> DEBUG   : starting all dmraids on drives ['sda', 'sdb']
> DEBUG   : scanning for dmraid on drives ['sda', 'sdb']
> DEBUG   : got raidset <block.device.RaidSet instance at 0xb7d173ac> (sda 
> sdb)
> DEBUG   :   valid: True found_devs: 2 total_devs: 2
> DEBUG   : adding 
> mapper/ddf1_44656c6c202020201028001510281f0332b494acaae97d7b to isys cache
> DEBUG   : adding 
> mapper/ddf1_44656c6c202020201028001510281f0332b494acaae97d7b to isys 
> cache()
> DEBUG   : adding sda to dm cache
> DEBUG   : removing sda from isys cache
> DEBUG   : adding sdb to dm cache
> DEBUG   : removing sdb from isys cache
>
> ## it detected dmraid?!?!
> -->>DEBUG   : starting raid <block.device.RaidSet instance at 
> 0xb7d173ac> with mknod=True
>
> DEBUG   : done starting dmraids.  Drivelist: 
> ['mapper/ddf1_44656c6c202020201028001510281f0332b494acaae97d7b']
> INFO    : Running kickstart %%pre script(s)
> INFO    : All kickstart %%pre script(s) have been run
>
> -- Correctly, does not detect dmraid. --
>
> DEBUG   : starting dmraids
> DEBUG   : self.driveList(): ['sda', 'sdb']
> DEBUG   : DiskSet.skippedDisks: []
> DEBUG   : DiskSet.skippedDisks: []
> DEBUG   : starting all dmraids on drives ['sda', 'sdb']
> DEBUG   : scanning for dmraid on drives ['sda', 'sdb']
> DEBUG   : done starting dmraids.  Drivelist: ['sda', 'sdb']
> INFO    : Running kickstart %%pre script(s)
> INFO    : All kickstart %%pre script(s) have been run
>
> In the meantime, to get around this issue I just commented out all calls 
> to startDmRaid().  I will continue to decipher the code (anaconda, 
> python-pyblock, dmraid-devel) , but I wanted to ask here first.
>
> Thanks,
> Christopher
>
> Anaconda-devel-list mailing list
> [e-mail]
> https://www.redhat.com/mailman/listinfo/anaconda-devel-list

[Niels de Vos]

On Fri, 2007-10-26 at 14:37 -0400, Christopher Boumenot wrote:
> Greetings,
>
> We are seeing an issue with anaconda-11.1.2.36 (CentOS 5.0), where it 
> incorrectly detects dmraid on our Dell 2900s (and, 1950s, 2950s).  (We 
> were able to track it down to that issue based on the message "Specified 
> nonexistent disk sda in partition command," and /tmp/anaconda.log.)  I 
> am not sure why Anaconda detects dmraid.

Hi,

anaconda supports the parameter 'nodmraid'. Maybe this helps.

Good luck,
Niels