Cool,

I've found a problem in the opkg-munge-* packages. I've just updated it. 
(V1.0.3) Please use that if possible. If not, just before step 7 do a service 
munge start on the head and on all nodes.
I'm currently checking that it's now ok. (it's a wired voodoo problem... munge 
is one of the easiest thing to configure and the one that gave me most 
problems... I need holidays ;-))

PS: don't forget to fix you /var/lib/systemimager/scripts/*.master line 190 if 
it has been generated with pre 4.2.0-0.92 version.

Best regards,

Olivier.
--
   Olivier LAHAYE
   CEA DRT/LIST/DCSI/DIR
________________________________
De : Costel Seitan [csei...@slb.com]
Date d'envoi : mercredi 27 février 2013 13:48
À : LAHAYE Olivier
Cc: oscar-users@lists.sourceforge.net
Objet : RE: RE :OSCAR on CentOS-6: how to test.(Warning, experimental) WAS:Re: 
OSCAR on CentOS-6: bootstrap successfull using online repository.

Hi Olivier,

I updated my packages with yum and even installed munge packages.  Now
the right version of system imager is installed.
I see the official repo has been updated with your last packages built on 
yesterday so I used that one.

I will transfer the last pre- and post-install scripts to 
/var/lib/systemimager/scripts/ and then
try again the Oscar wizard.

I will check the dhcpd.conf as you recommended.


Thank you for your help,
Kind regards,
Costel  SEITAN

From: LAHAYE Olivier [mailto:olivier.lah...@cea.fr]
Sent: Wednesday, February 27, 2013 11:42 AM
To: Costel Seitan
Cc: oscar-users@lists.sourceforge.net
Subject: RE : RE :OSCAR on CentOS-6: how to test.(Warning, experimental) 
WAS:Re: OSCAR on CentOS-6: bootstrap successfull using online repository.

Hi,

systemimager release 4.2.0-0.92 fixes a bug in the master script and adds some 
drivers to the kernel.
If your image has been generated with the 0.91 release, you need to update the 
master script: line 190.
You need to replace egrep '$DISK0' with egrep "$DISK0".
/tftpboot/distro/rhel-6-x86_64 is already used by yume (not yum) as it is 
defined in /tftpboot/distro/rhel-6-x86_64.url file. the 
/etc/oscar/supported_distros.txt is used by the setup-distro to generate 
/tftpboot/{oscar,distro}/rhel-6-x86_64.url files. The local repo are 
automaticaly added. (Though, they are often empty except if you did built oscar 
yourself).

Most important, you need the preinstall script that modprobe some disk 
controlers. You also need the post-install scripts that sets up the grub 
bootloader and the network. The reason for those post-install scriipts is that 
systemconfig is not yet ported to the newer linux architecture (networkmanager, 
grub/grub2, systemd, ...)

Latest tips:
- The /etc/dhcp/dhcpd.conf is badly generated by oscar: the option routers is 
bad, please update it. (need a fix into oscar)
- You can add dns info if you want to have access to the internet from your 
nodes (yum update for example).
    option domain-name "you.dns.domain.com";
    option domain-name-servers 192.168.1.1, 192.168.1.2;
=> Then restart the dhcpd service before before trying to deploy the nodes.

Today, I completed the oscar installation up to step 6 starting from a blank 
wiped out VM (both server and node). Step 7 failed because munge was not 
started, I have to investigate, surely a easy to fix problem.

Happy testing,

Best regards,

Olivier.
--
   Olivier LAHAYE
   CEA DRT/LIST/DCSI/DIR
________________________________
De : Costel Seitan [csei...@slb.com]
Date d'envoi : mercredi 27 février 2013 10:29
À : LAHAYE Olivier
Cc: oscar-users@lists.sourceforge.net
Objet : RE: RE :OSCAR on CentOS-6: how to test.(Warning, experimental) WAS:Re: 
OSCAR on CentOS-6: bootstrap successfull using online repository.
Hi Olivier,

Just to let you know my OSCAR image was built successfully this night. Do you 
believe I can add
the local repository (/tftpboot/distro/rhel-6-x86_64) before epel and remote 
CentOS mirrors
in order to accelerate things or is not a good idea?

I see the systemimager installed is version 0.91.  I don’t know if it will work 
but I can give it a try.
I would execute the post-install script if needed.

Concerning the repo I thought of the same issue about the limited bandwidth.
It took longtime but it finally did the job.

I disabled iptables, SELinux (and maybe IPV6 too) from the beginning. My 
previous experience
with OSCAR and Beowulf clusters was of good help.

I will also use Torque because I will certainly do not have time to setup SGE.

Keep in touch,
Costel SEITAN

From: LAHAYE Olivier [mailto:olivier.lah...@cea.fr]
Sent: Tuesday, February 26, 2013 6:55 PM
To: Costel Seitan
Cc: oscar-users@lists.sourceforge.net
Subject: RE :OSCAR on CentOS-6: how to test.(Warning, experimental) WAS:Re: 
OSCAR on CentOS-6: bootstrap successfull using online repository.

Hi Costel,

I've reupdated my repository.

- I've hopefully fixed all issues with systeminstaller generating bad master 
script. I've also added a few drivers (mainly network and scsi (vmware at 
least)). I hope to have a far better systemimager later, but for the moment, 
this somewhat old rpm should do the trick.

- I've updated a few oscar packages, and munge now should show up, and some 
minor problems should be fixed as well.

- You are unfortunately hitting my repository problem: it's a non-pro web space 
free of charge from my ISP and thus the bandwidth is limited. That's why it's a 
"non-official" repo. The bandwidth and the size are limited due to my ISP 
contract that says I need to purchase pro-space if I need bandwidth.
  Anyway, keep trying, it should work in the end. Try to avoid 5PM to 9PM 
French time and it should work.

- regarding openmpi, I'll give a look, it's strange......You can disable all 
things that fails of course. I've also noted a tiny bug ion the torque package 
post install (the alternatives are not properly set for qsub 
(/etc/alternatives/qsub points at the wrong place). I'll fix that ASAP. It 
should not prevent torque from installing and working though.

- regarding the image building, I had the same problem today, some epel mirrors 
were very slow....
  Note that the systemimager that corrects the master script generation bug is 
the 0.92 release.
  Note also that you may require the following pre and post install scripts to 
circumvent current systemimager issues  
http://olivier.lahaye1.free.fr/RPMS/extra/si_scripts.tar.bz2
  I've noted that packman crashes when the package download fails. I'll add 
this on my todo list.

- regarding SGE, the oscar materials are completely obsoletes. now the system 
is called gridengine and there are some rpms available. I think that we need to 
rewrite the whole stuff in oscar. I'll give a look when oscar for centos-6 will 
be running fine. I hope to be able to use the officials gridengine rpms so I 
can avoid maintaining the package. If it's possible, then we would only have to 
update the opkgs.
For the moment, only torque + maui works in oscar.....(you can use gridengine 
on an oscar scluster, but you'll have to configure the whole stuff by hand. 
(for the moment)

IMPORTANT: I forgot to tell also that you should disable iptables and SELinux 
(or set it to permissive).
I'm not sure, but if my memory is correct, I had issues in the past when ipv6 
was enabled and I can't tell if I fixed the problem or not. The issue was when 
oscar was parsing the /etc/hosts grepping for localhost or something like that.

Today, I was able to reach again oscar step 7 successfully.(still minor issues 
in step 8 (testing))
Now, I've formatted my VMS and started again from scratch to check the 
remaining issues. Now building the oscarnode image. No errors so far.

See you tomorrow.

Best regards.
--
   Olivier LAHAYE
   CEA DRT/LIST/DCSI/DIR
________________________________
De : Costel Seitan [csei...@slb.com]
Date d'envoi : mardi 26 février 2013 18:02
À : LAHAYE Olivier
Cc: oscar-users@lists.sourceforge.net
Objet : RE: [Oscar-users] OSCAR on CentOS-6: how to test.(Warning, 
experimental) WAS:Re: OSCAR on CentOS-6: bootstrap successfull using online 
repository.
Hello,

Today I did a big step forward.
I did what you suggested and followed your list Olivier as you described into 
the previous posts.

NOTE: The available repositories were:
http://olivier.lahaye1.free.fr/OSCAR/repos/unstable/rhel-6-x86_64/, 
http://download.fedoraproject.org/pub/epel/6/x86_64, 
http://mirror.centos.org/centos/6/updates/x86_64/, 
http://mirror.centos.org/centos/6/extras/x86_64/, 
http://mirror.centos.org/centos/6/os/x86_64/
(I modified the supported_distros.txt and the url was recreated with 
http://olivier.lahaye1.free.fr/OSCAR/repos/unstable/rhel-6-x86_64/)

The Oscar installed OK and I successfully launched (at step 7)

oscar_wizard -d install

I selected the same packages as the one in your list.
NOTE: I do not see munge into the list of packages and it looks like during the 
installation of the selected packages it downloads munge from CentOs 6 
repository.

I configured the packages successfully.
It first failed at the step “Install OSCAR Server Packages” with the messages:


http://olivier.lahaye1.free.fr/OSCAR/repos/unstable/rhel-6-x86_64/openmpi-oscar-runtime-1.6.3-1.x86_64.rpm:
 [Errno 12] Timeout on 
http://olivier.lahaye1.free.fr/OSCAR/repos/unstable/rhel-6-x86_64/openmpi-oscar-runtime-1.6.3-1.x86_64.rpm:
 (28, 'Operation too slow. Less than 1 bytes/sec transfered the last 30 
seconds')
Trying other mirror.


Error Downloading Packages:
  openmpi-oscar-runtime-1.6.3-1.x86_64: failure: 
openmpi-oscar-runtime-1.6.3-1.x86_64.rpm from unstable_rhel-6-x86_64: [Errno 
256] No more mirrors to try.

ERROR during execution 256
Use of uninitialized value $output[0] in join or string at 
/usr/share/perl5/vendor_perl/OSCAR/PackMan.pm line 656.
Error occured during smart_install (-1):
at /usr/bin/install_server line 121


I noticed it is always the same .rpm file concerning openmpi. I therefore 
removed openmpi from the distribution and OSCAR packges installed correctly.
However when launching “Create a SIS Image” it took longtime with messages like:

http://download.fedoraproject.org/pub/epel/6/x86_64/repodata/repomd.xml: [Errno 
14] PYCURL ERROR 22 - "The requested URL returned error: 503"
Trying other mirror.
http://download.fedoraproject.org/pub/epel/6/x86_64/repodata/repomd.xml: [Errno 
12] Timeout on 
http://download.fedoraproject.org/pub/epel/6/x86_64/repodata/repomd.xml: (28, 
'Operation too slow. Less than 1 bytes/sec transfered the last 30 seconds')

Indeed, I noticed that some of fedora repositories were hard to reach today.  
After 1 hour the SIS image building is still in progress.
I will let it run till tomorrow morning and I hope that the build of SIS image 
will finally go smoothly so I can pass to the first node installation.

NOTE: I suppose that SGE is somehow decommissioned or maybe you prefer Torque.  
I do not have experience with Torque but it looks similar.


Kind Regards,
Costel SEITAN




From: LAHAYE Olivier [mailto:olivier.lah...@cea.fr]
Sent: Monday, February 25, 2013 7:50 PM
To: oscar-users@lists.sourceforge.net
Subject: [Oscar-users] OSCAR on CentOS-6: how to test.(Warning, experimental) 
WAS:Re: OSCAR on CentOS-6: bootstrap successfull using online repository.

Costel,

If you want to test, you can update the /etc/oscar/supported_distros.txt and 
replace the oscar:http://official url with 
http://olivier.lahaye1.free.fr/OSCAR/repos/unstable/rhel-6-x86_64/

alternative method: after the bootstrapping:

in /tftpboot/oscar/rhel-6-x86_64.url, replace the official url with 
http://olivier.lahaye1.free.fr/OSCAR/repos/unstable/rhel-6-x86_64/ Note that 
the url is ofter rewritten. The prefered method is the 1st one.

So:
1: rpm -Uvh 
http://olivier.lahaye1.free.fr/OSCAR/repos/unstable/rhel-6-x86_64/oscar-release-6.1.2r9625-1.noarch.rpm
2: yum install oscar
3: vim /etc/oscar/supported_distros.txt
4: oscar-config --setup-distro centos-6-x86_64
5: verify that /tftpboot/oscar/rhel-6-x86_64.url contains the correct url
6: oscar-config --bootstrap
7: oscar_wizard -d install
And so on.

WARNING: This repo is not official and sometimes it's broken. be warned of that 
when using it for testing.

Best regards,

Olivier.

--
   Olivier LAHAYE
   CEA DRT/LIST/DCSI/DIR
________________________________
________________________________
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_feb
_______________________________________________
Oscar-users mailing list
Oscar-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oscar-users

Reply via email to