CentOS - Apr 2015 - CentOS6.6 installation failure

I have been installing CentOS for years using PXE, kickstart and NFS, but the
procedure fails with CentOS 6.6 for some reason.

A PXE-initiated, manual install goes through all the setup steps, and once
rpm installation has started, it dies after a few packages:

anaconda 13.21.229 exception report
Traceback (most recent call first):
  File "/usr/lib/anaconda/yuminstall.py", line 1006, in _handleFailure
    os.unlink(package.localPkg())
  File "/usr/lib/anaconda/yuminstall.py", line 196, in callback
    self.ayum._handleFailure(po, trynumber)
OSError: [Errno 30] Read-only file system:
'/mnt/source/Packages/mailcap-2.1.31-2.el6.noarch.rpm'

When a kickstart file is used instead, the installation dies at a different
package.

I have tried providing the OS installation files from different machines,
checked package integrity etc., but cannot figure out what is happening here.

The anaconda traceback file also has

12:19:59,354 WARNING : Failed to get file:///mnt/source/Packages/mailcap-2.1.31-
2.el6.noarch.rpm from mirror 1/1, or downloaded file is corrupt
12:19:59,354 WARNING : package download failure, retrying automatically

That said, installing 6.4 off the same NFS server works just fine.

On 04/20/2015 07:21 AM, lhecking at users.sourceforge.net
wrote:
> 
> I have been installing CentOS for years using PXE, kickstart and NFS, but
the
> procedure fails with CentOS 6.6 for some reason.
> 
> A PXE-initiated, manual install goes through all the setup steps, and once
> rpm installation has started, it dies after a few packages:
> 
> anaconda 13.21.229 exception report
> Traceback (most recent call first):
>   File "/usr/lib/anaconda/yuminstall.py", line 1006, in
_handleFailure
>     os.unlink(package.localPkg())
>   File "/usr/lib/anaconda/yuminstall.py", line 196, in callback
>     self.ayum._handleFailure(po, trynumber)
> OSError: [Errno 30] Read-only file system:
'/mnt/source/Packages/mailcap-2.1.31-2.el6.noarch.rpm'
> 
> When a kickstart file is used instead, the installation dies at a different
> package.
> 
> I have tried providing the OS installation files from different machines,
> checked package integrity etc., but cannot figure out what is happening
here.
> 
> The anaconda traceback file also has
> 
> 12:19:59,354 WARNING : Failed to get
file:///mnt/source/Packages/mailcap-2.1.31-
> 2.el6.noarch.rpm from mirror 1/1, or downloaded file is corrupt
> 12:19:59,354 WARNING : package download failure, retrying automatically
> 
> That said, installing 6.4 off the same NFS server works just fine.
I would say to eliminate issues, the first thing to make sure is that
your mailcap package is, indeed, good.




-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: OpenPGP digital signature
URL:
<http://lists.centos.org/pipermail/centos/attachments/20150420/56b558b3/attachment-0001.sig>

> I would say to eliminate issues, the first thing to make sure is that
> your mailcap package is, indeed, good.
 Summary: it's a good package, but the wrong package for the location.

 I downloaded the package from a different mirror, and the install proceeds
 past it. However, now it stops at redhat-logos, with the same error

 An interesting observation: I now have several copies of the mailcap package
 on my system. When I compare md5sums, I see this:

 CentOS 6.4 tree:

029213d1c7b79c0dfa4f6d8063dbdce6  mailcap-2.1.31-2.el6.noarch.rpm (i386)
65e231403f819bea3dce22adee5cb373  mailcap-2.1.31-2.el6.noarch.rpm (x86_64)

 CentOS 6.6 tree, where the x86_64 version was replaced with a fresh copy:

029213d1c7b79c0dfa4f6d8063dbdce6  mailcap-2.1.31-2.el6.noarch.rpm (i386)
029213d1c7b79c0dfa4f6d8063dbdce6  mailcap-2.1.31-2.el6.noarch.rpm (x86_64)
65e231403f819bea3dce22adee5cb373  mailcap-2.1.31-2.el6.noarch.rpm.dist
(previous)

 The packages in 6.4 are different for each arch, but in 6.6 they are the same.
 It appears this relates to how precisely you guys build a distribution?

 IIRC I took a shortcut when upgrading my repositories from 6.x -> 6.x+1
 by pre-populating the directories with the previous version to save some b/w.
 This could be the reason for what I see here.

 Now, if there were checksum files for everything, I knew which rpms I need
 to download again, but there aren't. Looks like the noarch packages are a
 good start, though.