[Rocks-Discuss] Cannot retrieve repository metadata (repomd.xml)

Arild Sæther

unread,

Mar 17, 2011, 8:18:49 AM3/17/11

to npaci-rocks...@sdsc.edu

Hi,

I am struggling with Rocks 5.4 and making it install my compute nodes. Currently it gets through the pre-install scripts fine, but then stops as soon as the RPM installation is about to begin with the following error:

Reading package metadata: Cannot retrieve repository metadata (repomd.xml) for repository anaconda-base-201101061125.x86_64. Please verify its path and try again.

I can locate a "repomd.xml" file in several locations:

/state/partition1/rocks/install/rocks-dist/x86_64/RedHat/RPMS/repomd.xml
/state/partition1/rocks/install/rocks-dist/x86_64/repodata/repomd.xml
/state/partition1/rocks/install/rolls/Red_Hat_Enterprise_Linux_Server_5.4/5.4/x86_64/RedHat/RPMS/repomd.xml
/state/partition1/rocks/install/tmpYe7SGS/x86_64/RedHat/RPMS/repomd.xml
/state/partition1/rocks/install/tmpYe7SGS/x86_64/repodata/repomd.xml

But apparently, neither can be found by the node (or neither are suitable). The top entry in this list links to entry number 2, which seems to have the correct time stamp (ie. It is from the last time I ran "rocks create distro").

I should also mention that when I rebuild the distro, I have to manually add a link from /var/www/html/install/rocks-dist/x86_64/images/product.img to /state/partition1/rocks/install/rolls/kernel/5.4/x86_64/images/product.img to make the system even get to the pre-installation part. As I see it, there is a chance there are things that Rocks should be doing (like creating links) during the rocks create distro phase which it is not doing.

However, I am having a hard time finding out where it is trying to grab repomd.xml from in the first place. Can anyone help?

Best regard,
Arild

-------------------------------------------------------------------
The information contained in this message may be CONFIDENTIAL and is
intended for the addressee only. Any unauthorised use, dissemination of the
information or copying of this message is prohibited. If you are not the
addressee, please notify the sender immediately by return e-mail and delete
this message.
Thank you
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.sdsc.edu/pipermail/npaci-rocks-discussion/attachments/20110317/2f86b527/attachment.html

Rick Wagner

unread,

Mar 17, 2011, 11:35:13 AM3/17/11

to Discussion of Rocks Clusters

Hi Arild,

On Mar 17, 2011, at 5:18 AM, Arild Sæther wrote:

> Hi,
>
> I am struggling with Rocks 5.4 and making it install my compute nodes. Currently it gets through the pre-install scripts fine, but then stops as soon as the RPM installation is about to begin with the following error:
>
> Reading package metadata: Cannot retrieve repository metadata (repomd.xml) for repository anaconda-base-201101061125.x86_64. Please verify its path and try again.
>
> I can locate a "repomd.xml" file in several locations:
>
> /state/partition1/rocks/install/rocks-dist/x86_64/RedHat/RPMS/repomd.xml
> /state/partition1/rocks/install/rocks-dist/x86_64/repodata/repomd.xml
> /state/partition1/rocks/install/rolls/Red_Hat_Enterprise_Linux_Server_5.4/5.4/x86_64/RedHat/RPMS/repomd.xml
> /state/partition1/rocks/install/tmpYe7SGS/x86_64/RedHat/RPMS/repomd.xml
> /state/partition1/rocks/install/tmpYe7SGS/x86_64/repodata/repomd.xml
>
> But apparently, neither can be found by the node (or neither are suitable). The top entry in this list links to entry number 2, which seems to have the correct time stamp (ie. It is from the last time I ran "rocks create distro").
>
> I should also mention that when I rebuild the distro, I have to manually add a link from /var/www/html/install/rocks-dist/x86_64/images/product.img to /state/partition1/rocks/install/rolls/kernel/5.4/x86_64/images/product.img to make the system even get to the pre-installation part. As I see it, there is a chance there are things that Rocks should be doing (like creating links) during the rocks create distro phase which it is not doing.
>
> However, I am having a hard time finding out where it is trying to grab repomd.xml from in the first place. Can anyone help?

Can you post a list of the rolls you're using (from 'rocks list roll'), and have you made any customizations, e.g., created an extend-compute file?

Thanks,
Rick

Greg Bruno

unread,

Mar 17, 2011, 11:40:55 AM3/17/11

to Discussion of Rocks Clusters

2011/3/17 Arild Sæther <ari...@statoil.com>:

> Hi,
>
> I am struggling with Rocks 5.4 and making it install my compute nodes. Currently it gets through the pre-install scripts fine, but then stops as soon as the RPM installation is about to begin with the following error:
>
> Reading package metadata: Cannot retrieve repository metadata (repomd.xml) for repository anaconda-base-201101061125.x86_64. Please verify its path and try again.
>
> I can locate a "repomd.xml" file in several locations:
>
> /state/partition1/rocks/install/rocks-dist/x86_64/RedHat/RPMS/repomd.xml
> /state/partition1/rocks/install/rocks-dist/x86_64/repodata/repomd.xml
> /state/partition1/rocks/install/rolls/Red_Hat_Enterprise_Linux_Server_5.4/5.4/x86_64/RedHat/RPMS/repomd.xml
> /state/partition1/rocks/install/tmpYe7SGS/x86_64/RedHat/RPMS/repomd.xml
> /state/partition1/rocks/install/tmpYe7SGS/x86_64/repodata/repomd.xml
>
> But apparently, neither can be found by the node (or neither are suitable). The top entry in this list links to entry number 2, which seems to have the correct time stamp (ie. It is from the last time I ran "rocks create distro").
>
> I should also mention that when I rebuild the distro, I have to manually add a link from /var/www/html/install/rocks-dist/x86_64/images/product.img to /state/partition1/rocks/install/rolls/kernel/5.4/x86_64/images/product.img to make the system even get to the pre-installation part. As I see it, there is a chance there are things that Rocks should be doing (like creating links) during the rocks create distro phase which it is not doing.
>
> However, I am having a hard time finding out where it is trying to grab repomd.xml from in the first place. Can anyone help?

It sounds like the frontend did not install correctly.

You should verify the MD5 checksums of the ISOs you downloaded, reburn
the ISOs to CDs (or DVDs), then reinstall the frontend.

Also, since you used Red Hat Enterprise Linux Server v5.4 as a
substitue for the OS rolls, make sure you supply all the RHEL CDs when
you install the frontend.

- gb

Arild Sæther

unread,

Mar 18, 2011, 4:59:39 AM3/18/11

to Discussion of Rocks Clusters

Hi Rick,

Thank you for your prompt reply. Here is the output of rocks list roll:

NAME VERSION ARCH ENABLED
kernel: 5.4 x86_64 yes
base: 5.4 x86_64 yes
hpc: 5.4 x86_64 yes
service-pack: 5.4.2 x86_64 yes
web-server: 5.4 x86_64 yes
Red_Hat_Enterprise_Linux_Server_5.4: 5.4 x86_64 yes
Red_Hat_Enterprise_Linux_Server_5.4: 5.4 x86_64 yes
Red_Hat_Enterprise_Linux_Server_5.4: 5.4 x86_64 yes
Red_Hat_Enterprise_Linux_Server_5.4: 5.4 x86_64 yes
Red_Hat_Enterprise_Linux_Server_5.4: 5.4 x86_64 yes
None: 5.4 None yes

We have indeed created an extend-file for the compute nodes. Is that likely to break things the way we see it happening here. I should mention that it is based on an extend-file which works fine in 5.3, but I may of course have mangled it without being able to spot it.

Best regards,
Arild Sæther

Sudarshan Wadkar

unread,

Mar 18, 2011, 6:37:55 AM3/18/11

to Discussion of Rocks Clusters

make sure your extend file is without any xml error

# xmllinit -noout <name_of_extend_file>
should return nothing, if it shows errors then you have problems

-Sudarshan Wadkar
System Administrator
High Performance Computing Center
IIT Bombay, Powai, Mumbai 400 076

2011/3/18 Arild Sæther <ari...@statoil.com>:

--
~$udhi
"Success is getting what you want. Happiness is wanting what you get."
- Dale Carnegie
"It's always our decision who we are"
- Robert Solomon in Waking Life

Anoop Rajendra

unread,

Mar 18, 2011, 9:40:01 AM3/18/11

to Discussion of Rocks Clusters

Looks like you have a roll called the "None" roll?

That looks suspicious. Are you certain that that's supposed to be there?

Can you disable the roll

# rocks disable roll None

re-create the distribution

# cd /export/rocks/install
# rm -rf rocks-dist

and then try to re-install?

-a
Sent from my iPhone

Arild Sæther

unread,

Mar 23, 2011, 8:38:45 AM3/23/11

to Discussion of Rocks Clusters

Hi,

Thanks for all the help. In the end, I reinstalled, and it seemed the first installation hadn't gone great. The problem went away, but I found other ones. Always something :)

Arild

Sudarshan Wadkar

unread,

Mar 23, 2011, 9:09:05 AM3/23/11

to Discussion of Rocks Clusters

Feel free to communicate. The list is always happy to help :)

2011/3/23 Arild Sæther <ari...@statoil.com>:

> Hi,
>
> Thanks for all the help. In the end, I reinstalled, and it seemed the first installation hadn't gone great. The problem went away, but I found other ones. Always something :)
>
> Arild
>

--
-Sudarshan Wadkar
Research Assistant & System Administrator

High Performance Computing Center
IIT Bombay, Powai, Mumbai 400 076

"Success is getting what you want. Happiness is wanting what you get."

- Dale Carnegie
"It's always our decision who we are"
- Robert Solomon in Waking Life

"The Truth is The Truth, so all you can do is live with it."
- $udhi :)

Jack Fowler

unread,

Apr 12, 2011, 4:22:34 PM4/12/11

to Discussion of Rocks Clusters

Hi All,
I'm trying to install a login node and am getting the same error as reported in an earlier thread:

> Reading package metadata: Cannot retrieve repository metadata (repomd.xml) for repository
> anaconda-base-201101061125.x86_64. Please verify its path and try again.
>

I also have an extend-compute.xml file, here's the output of xmllint:

[root@brux2 nodes]# xmllint -noout extend-compute.xml
extend-compute.xml:50: parser error : Entity 'Kickstart_PublicHostname' not defined
wget http://&Kickstart_PublicHostname;/install/rocks-dist/x86_64/RedHat/
^
extend-compute.xml:98: parser error : Entity 'Kickstart_PublicHostname' not defined
wget -P /etc http://&Kickstart_PublicHostname;/install/rocks-dist/x86_64

I've installed new compute nodes with this file as recently as last week. I did make some edits, now
removed, but they didn't touch the above lines.
Any ideas?

thanks,
Jack

Anoop Rajendra

unread,

Apr 12, 2011, 4:35:00 PM4/12/11

to Discussion of Rocks Clusters

On Tue, Apr 12, 2011 at 1:22 PM, Jack Fowler <fow...@hep.brown.edu> wrote:
> Hi All,
> I'm trying to install a login node and am getting the same error as reported in an earlier thread:
>
>> Reading package metadata: Cannot retrieve repository metadata (repomd.xml) for repository
>> anaconda-base-201101061125.x86_64. Please verify its path and try again.
>>
>
> I also have an extend-compute.xml file, here's the output of xmllint:
>
> [root@brux2 nodes]# xmllint -noout extend-compute.xml
> extend-compute.xml:50: parser error : Entity 'Kickstart_PublicHostname' not defined
> wget http://&Kickstart_PublicHostname;/install/rocks-dist/x86_64/RedHat/
> ^
> extend-compute.xml:98: parser error : Entity 'Kickstart_PublicHostname' not defined
> wget -P /etc http://&Kickstart_PublicHostname;/install/rocks-dist/x86_64

This error occurs because the entities exist only in the context of
kickstart generation. xmllint is not run in this context.

What is the output of

# rocks list host profile <nodename>

-a

Jack Fowler

unread,

Apr 13, 2011, 9:50:28 AM4/13/11

to Discussion of Rocks Clusters

On 04/12/2011 04:35 PM, Anoop Rajendra wrote:
> On Tue, Apr 12, 2011 at 1:22 PM, Jack Fowler <fow...@hep.brown.edu> wrote:
>> Hi All,
>> I'm trying to install a login node and am getting the same error as reported in an earlier thread:
>>
>>> Reading package metadata: Cannot retrieve repository metadata (repomd.xml) for repository
>>> anaconda-base-201101061125.x86_64. Please verify its path and try again.
>>>
>>
>> I also have an extend-compute.xml file, here's the output of xmllint:
>>
>> [root@brux2 nodes]# xmllint -noout extend-compute.xml
>> extend-compute.xml:50: parser error : Entity 'Kickstart_PublicHostname' not defined
>> wget http://&Kickstart_PublicHostname;/install/rocks-dist/x86_64/RedHat/
>> ^
>> extend-compute.xml:98: parser error : Entity 'Kickstart_PublicHostname' not defined
>> wget -P /etc http://&Kickstart_PublicHostname;/install/rocks-dist/x86_64
>
> This error occurs because the entities exist only in the context of
> kickstart generation. xmllint is not run in this context.
>
> What is the output of
>
> # rocks list host profile <nodename>
>
> -a

Nodename? Headnode or target node? Using either produces a very long output. I can attach these if
you want?
thanks,
Jack

Anoop Rajendra

unread,

Apr 13, 2011, 3:22:27 PM4/13/11

to Discussion of Rocks Clusters

On Wed, Apr 13, 2011 at 6:50 AM, Jack Fowler <fow...@hep.brown.edu> wrote:
>
>
> On 04/12/2011 04:35 PM, Anoop Rajendra wrote:
>> On Tue, Apr 12, 2011 at 1:22 PM, Jack Fowler <fow...@hep.brown.edu> wrote:
>>> Hi All,
>>> I'm trying to install a login node and am getting the same error as reported in an earlier thread:
>>>
>>>> Reading package metadata: Cannot retrieve repository metadata (repomd.xml) for repository
>>>> anaconda-base-201101061125.x86_64. Please verify its path and try again.
>>>>
>>>
>>> I also have an extend-compute.xml file, here's the output of xmllint:
>>>
>>> [root@brux2 nodes]# xmllint -noout extend-compute.xml
>>> extend-compute.xml:50: parser error : Entity 'Kickstart_PublicHostname' not defined
>>> wget http://&Kickstart_PublicHostname;/install/rocks-dist/x86_64/RedHat/
>>> ^
>>> extend-compute.xml:98: parser error : Entity 'Kickstart_PublicHostname' not defined
>>> wget -P /etc http://&Kickstart_PublicHostname;/install/rocks-dist/x86_64
>>
>> This error occurs because the entities exist only in the context of
>> kickstart generation. xmllint is not run in this context.
>>
>> What is the output of
>>
>> # rocks list host profile <nodename>
>>
>> -a
>
> Nodename? Headnode or target node? Using either produces a very long output. I can attach these if
> you want?
> thanks,
> Jack

As long as the "rocks list host profile" command does not produce an
error, the changes you've made to extend-compute.xml should be OK.

That said, you still have to deal with the repomd.xml problem. Rebuild
the distribution and try you installation again.

# cd /export/rocks/install
# rm -rf rocks-dist

# rocks create distro

Re-run the installation

-a

Jack Fowler

unread,

Apr 13, 2011, 4:53:04 PM4/13/11

to Discussion of Rocks Clusters

OK, I did that and it still fails. Here are the exact error screens:

Unable to read package metadata.This may be due
to a missing repodata directory. Please ensure that
your install tree has been correctly generated.

Cannot retrieve repository metadata (repomd.xml)

for repository: anaconda-base-200912121608.x86_64
Please verify it's path and try again.

I click continue and:

Unable to read group information from repositories.
This is a problem with the generation of your install tree.

The next screen is a debug screen which asks to save to a file. I've done that and can attach it if
you want?

The last successful node install was on April 6 after which I ran yum update on all nodes. Could
that have caused the problem? Here are the updates:
Apr 06 16:43:00 Updated: tzdata-2011d-3.el5.x86_64
Apr 06 16:43:24 Updated: glibc-common-2.5-58.el5_6.2.x86_64
Apr 06 16:43:45 Updated: glibc-2.5-58.el5_6.2.x86_64
Apr 06 16:43:48 Updated: glibc-2.5-58.el5_6.2.i686
Apr 06 16:43:48 Updated: libselinux-1.33.4-5.7.el5.x86_64
Apr 06 16:43:48 Updated: libselinux-1.33.4-5.7.el5.i386
Apr 06 16:43:50 Updated: nscd-2.5-58.el5_6.2.x86_64
Apr 06 16:43:50 Updated: libselinux-python-1.33.4-5.7.el5.x86_64
Apr 06 16:43:50 Updated: libselinux-utils-1.33.4-5.7.el5.x86_64
Apr 06 16:43:52 Updated: glibc-headers-2.5-58.el5_6.2.x86_64
Apr 06 16:43:53 Updated: libselinux-devel-1.33.4-5.7.el5.i386
Apr 06 16:43:54 Updated: glibc-devel-2.5-58.el5_6.2.x86_64
Apr 06 16:43:55 Updated: glibc-devel-2.5-58.el5_6.2.i386
Apr 06 16:43:55 Updated: libselinux-devel-1.33.4-5.7.el5.x86_64
Apr 12 09:36:19 Updated: 12:libdhcp4client-3.0.5-23.el5_6.4.x86_64
Apr 12 09:36:20 Updated: 12:dhcp-3.0.5-23.el5_6.4.x86_64
Apr 12 09:36:21 Updated: 12:dhclient-3.0.5-23.el5_6.4.x86_64
Apr 12 09:36:25 Updated: 2:postfix-2.3.3-2.2.el5_6.x86_64
Apr 12 09:36:25 Updated: 12:dhcp-devel-3.0.5-23.el5_6.4.x86_64
Apr 12 09:36:25 Installed: php-IDNA_Convert-0.6.3-2.el5.noarch
Apr 12 09:36:26 Installed: php-simplepie-1.1.3-3.el5.noarch
Apr 12 09:36:31 Updated: wordpress-3.1-1.el5.noarch
Apr 12 09:36:32 Updated: 12:libdhcp4client-3.0.5-23.el5_6.4.i386
Apr 12 09:36:33 Updated: 12:libdhcp4client-devel-3.0.5-23.el5_6.4.x86_64
Apr 12 17:02:38 Updated: xorg-x11-server-utils-7.1-5.el5_6.1.x86_64

I'm at a loss why this isn't working. Any help is appreciated.
thanks,
Jack

Anoop Rajendra

unread,

Apr 14, 2011, 1:54:13 PM4/14/11

to Discussion of Rocks Clusters

Ah, a yum update.

The list is replete with posts about this. I'm afraid yum updates can
cause unpredictable behaviour, and can wreak havoc on the cluster.