I've had several unsuccessful attempts at installing 6.0.1 from
http://bear.csm.ornl.gov/repos/rhel-5-i386 on a 10 node cluster running
a fresh installation of Centos 5.2

The main issues are:-

1. configurator package is not installed automatically

2. At step 6 I get:
ERROR: Impossible to get the network interface id we have to use on
compute nodes (oscar_server, eth0) at /usr/bin/setup_pxe line 181.

3. After nodes are imaged and I try to run step 7 I get prompted to add
host key to known_hosts file and for root password which is not set.


I think that the second and third problems are connected to the fact
that the oscar mysql database is not being populated with all the
information required.
I'll try and summarise the problems I've had and include log extracts
where appropriate.

PROBLEM 1

Firstly, running  oscar-config --setup-distro centos-5-i386 sets the
file /tftpboot/oscar/rhel-5-i386.url to point to the unstable repository
http://bear.csm.ornl.gov/repos/unstable/rhel-5-i386 not the stable one.
I need to modify this by hand.
Then, when I run oscar-config --bootstrap command it fails to install
the configurator package: 

Prereqs status (/usr/share/oscar/prereqs/Configurator):
configurator: needs to be installed


[ http://bear.csm.ornl.gov/repos/rhel-5-i386
http://isoredirect.centos.org/centos/5/os/i386/ ]
Array: 2 element(s)
smartly installing: configurator
ERROR during execution 512
package configurator is not installed
!!! Some packages were NOT installed !!!
ERROR: impossible to install Configurator ( /usr/bin/install_prereq
--smart /usr/share/oscar/prereqs/Configurator).


Checking the repository explains this as the package simply isn't there
so I grabbed it from the unstable repository and installled it manually
with

yum localinstall ~/configurator-1.0-1.noarch.rpm

Bootstap then completes (after manually restarting mysql). Next I run
oscar_wizard install and am able to create an image with the following
errors

=============================================================================
== Running step 4 of the OSCAR wizard: Build OSCAR client image
=============================================================================

0: centos-5-i386
Selection: centos-5-i386
[add_image] Starting... 
Executing command: mksiimage -A --name oscarimage --distro centos-5-i386
--filename /usr/share/oscar/oscarsamples/rhel-5-i386.rpmlist --arch i386
--path /var/lib/systemimager/images/oscarimage   --force 
ERROR during execution 512
ERROR during execution 512
ERROR during execution 512
ERROR during execution 512
WARNING: Impossible to install glibc-kernheaders (1, )
ERROR during execution 512
WARNING: Impossible to install slocate (1, )
ERROR during execution 512
WARNING: Impossible to install utempter (1, )
Image build finished.
Added Disk Table for oscarimage based
on /usr/share/oscar/oscarsamples/scsi.gg.disk
Running: si_mkautoinstallscript -quiet --autodetect-disks -image
oscarimage -force -ip-assignment static -post-install reboot ... done



PROBLEM 2

At step 6 in oscar_wizard there seems to be an issue with the database
not knowing the compute nodes' network interface.

ERROR: Impossible to get the network interface id we have to use on
compute nodes (oscar_server, eth0) at /usr/bin/setup_pxe line 181
ERROR:  at /usr/lib/perl5/vendor_perl/5.8.8/OSCAR/GUI_MAC.pm line 784

Possibly related to this is the the warning received when I run
system-sanity

------------------------------------------------------
WARNING: A valid NIC must be specified for the cluster
private network.
Valid NICs: eth0, eth1

You tried to use: <None>.
This may be normal if this is the first time you 
execute OSCAR.
------------------------------------------------------

Also at step 5 I  always get  the error

error opening file:  /etc/c3.conf

The /tftpboot files for network boot were not created at step 6 until
about the seventh attempt at installing oscar.  When they were finally
created it was after I changed the file /etc/oscar/oscar.conf to include
ODA_TYPE = file rather than ODA_TYPE = db. This is a bad idea I guess
and may or may not have been the reason for the files to be finally
created.


PROBLEM 3

After finally persuading the wizard to report success at every step up
to 6, I network booted the nodes and imaged them with out obvious
problems, however when they boot there are messages about SSH RSA keys
being created (which shouldn't happen right?). When I try to run step 7
I get prompted to add host key to known_hosts file and for the root
password (which isn't set). Again, this shouldn't happen?

Inspecting the mysql oscar database seems to reveal the problem i.e
missing information about the nodes.

mysql> use oscar
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A

Database changed
mysql> select * from Nodes;
+------------+---------+-----------+----------------+------+--------------+---------------------+----+----------+-----------+--------------+------+------+-------+---------+
| cluster_id | cpu_num | cpu_speed | dns_domain     | fqdn | group_name
| hostname            | id | image_id | installer | name         | ram
| swap | units | virtual |
+------------+---------+-----------+----------------+------+--------------+---------------------+----+----------+-----------+--------------+------+------+-------+---------+
|          1 |    NULL | NULL      | oscardomain.gg | NULL |
oscar_server | head.oscardomain.gg |  1 |        0 | NULL      |
oscar_server | NULL | NULL | NULL  | NULL    | 
+------------+---------+-----------+----------------+------+--------------+---------------------+----+----------+-----------+--------------+------+------+-------+---------+
1 row in set (0.00 sec)

mysql> select * from Nics;
Empty set (0.00 sec)

I can set the root passwords on the nodes by booting them in single user
mode and running pwconv then passwd. Step 7 will then report success
(after typing in the password 9 times, one for each node) . I can not
run test cluster though and I get the error:

Use of uninitialized value in concatenation (.) or string
at /usr/bin/oscar_wizard line 879.

This would appear to be because $OSCAR_HOME is not set which it should
be?


Thanks for looking, I hope someone has some suggestions.
------------------------------------------------------------------------------
_______________________________________________
Oscar-users mailing list
Oscar-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oscar-users

Reply via email to