[CentOS] anaconda in centOS 4 fails to read md raid arrays

[Feizhou]

Johnny Hughes wrote:
> On Tue, 2005-06-07 at 15:12 +0800, Feizhou wrote:
> 
>>Johnny Hughes wrote:
>>
>>>On Mon, 2005-06-06 at 05:35 -0500, Johnny Hughes wrote:
>>>
>>>
>>>>On Mon, 2005-06-06 at 16:05 +0800, Feizhou wrote:
>>>>
>>>>
>>>>>Has anybody encountered this problem?
>>>>>
>>>>>I have a box that has FC2 installed. I wanted to trash the FC2 
>>>>>installation and install CentOS 4.
>>>>>
>>>>>I have a pxeboot/dhcp/kickstart environment and so I tried automatic 
>>>>>disk partitioning (clear all partitions and then create new ones), 
>>>>>manual disk partitioning but everything else is automated and finally, 
>>>>>zero automation, just a manual installation through VNC.
>>>>>
>>>>>Once I reach a part that touchs disk, the process will stop and end up with:
>>>>>
>>>>> Traceback (most recent call last):
>>>>>  File "/var/tmp/anaconda-10.1.1.13//usr/lib/anaconda/gui.py", line 
>>>>>789, in nextClicked
>>>>>    self.dispatch.gotoNext()
>>>>>  File "/var/tmp/anaconda-10.1.1.13//usr/lib/anaconda/dispatch.py", 
>>>>>line 171, in gotoNext
>>>>>    self.moveStep()
>>>>>  File "/var/tmp/anaconda-10.1.1.13//usr/lib/anaconda/dispatch.py", 
>>>>>line 239, in moveStep
>>>>>    rc = apply(func, self.bindArgs(args))
>>>>>  File "/var/tmp/anaconda-10.1.1.13//usr/lib/anaconda/partitioning.py", 
>>>>>line 54, in partitionObjectsInitialize
>>>>>    partitions.setFromDisk(diskset)
>>>>>  File "/var/tmp/anaconda-10.1.1.13//usr/lib/anaconda/partitions.py", 
>>>>>line 140, in setFromDisk
>>>>>    diskset.startAllRaid()
>>>>>  File "/var/tmp/anaconda-10.1.1.13//usr/lib/anaconda/partedUtils.py", 
>>>>>line 542, in startAllRaid
>>>>>    DiskSet.mdList.extend(raid.startAllRaid(driveList))
>>>>>  File "/var/tmp/anaconda-10.1.1.13//usr/lib/anaconda/raid.py", line 
>>>>>106, in startAllRaid
>>>>>    mdList = scanForRaid(driveList)
>>>>>  File "/var/tmp/anaconda-10.1.1.13//usr/lib/anaconda/raid.py", line 
>>>>>56, in scanForRaid
>>>>>    (major, minor, raidSet, level, nrDisks, totalDisks, mdMinor) =\
>>>>>  File "/usr/lib/anaconda/isys.py", line 97, in raidsb
>>>>>    return raidsbFromDevice("/tmp/md")
>>>>>  File "/usr/lib/anaconda/isys.py", line 103, in raidsbFromDevice
>>>>>    rc = _isys.getraidsb(fd)
>>>>>SystemError: (5, 'Input/output error')
>>>>
>>>>I had a similar problem with a ppc install on RHEL. I think (but I am
>>>>not 100% positive) that it is a LVM1 / LVM2 compatibility issue. 
>>>>
>>>>What I had to do was use fdisk and wipe the partitions and then I
>>>>recreated them in anaconda using the raid and lvm tools in anaconda.  
>>>>
>>>>I have never actually had that problem on CentOS-4, but it is the same
>>>>code, so I'm sure it happens in C4 too.
>>>>_______________________________________________
>>>
>>>
>>>BTW ... It might also be either this:
>>>
>>>https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=149220
>>>
>>>or this
>>>
>>>https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=144827
>>
>>No LVM. I therefore do not believe what I encountered is related to 
>>those at all. The traceback is rather clear that this is a RAID related 
>>issue.
> 
> 
> If you tried the default partition install, it creates LVM volumes.

No. I don't. The traceback occured BEFORE I could get Disk Druid. This 
happens when I click the button for Disk Druid in the partition setup 
process.
> Which is why I suspected that it might be one of those two bugs.  I had
> a similar issue with RHEL ppc that was also RAID only related, so it is
> probably different.

Or the same.
> 
> We will have update one and a new version of anaconda very soon (after
> upstream provider releases update 1) ... so maybe you can try with that
> one after CentOS-4.1 is released.

I guess you guys don't fix things inherited from Redhat then? You wait 
for them to fix them?
> 
> If you can access the alt-f2 console, you can run fdisk manually.  I
> don't know how to do that via VNC or a serial console though.

VNC would have been possible...it was with previous versions of anaconda 
until Redhat decided to remove that option. RH 7.1 GUI installation had 
it for example.