hi Costel, thanks for your support :-)
For EPEL timeouts issue, you can create a mirror on your oscar server or on an
ftp server on a simple machine on your network.
You just need to download recursively the epel repository (RPMS + rpocache +
repodata)
Then, in the /etc/yum.repos.d, just edit the epel.repo: comment the mirrorlist
line and add a line baseurl with the same syntax.
If the repo is local, I think that you can use
file:///tftpboot/epel/$releasever/$basearch/ (note: there are 3 slashes)
You also have to update /etc/oscar/supported_distros.txt (used by yume thus the
image creation process)
So you have to sync ftp://epel_mirror_server/path_to_epel/epel/6/x86_64 to
/tftpboot/epel/6/x86_64 and it should work.
You can use the mirror software for that.
<ftp://<centos_mirror>/<path><ftp://<centos_mirror>/<path>
As for your install, the error saying the user cancelled the process is an
erroneous message. It shows up when the image install fails. I didn't had time
to investigate on that (added to my todo list, hard to find IMHO).
As for the rsync, I think that rsyncd should be disabled and
/etc/init.d/systemimager-server-rsyncd should be enabled by STEP6 enable
install mode button.
In fact you should have the following 2 processes:
tail -n 0 --follow=name /var/log/systemimager/rsyncd
/usr/bin/rsync --daemon --confog=/etc/systemimager/rsyncd.conf
Though I understand that you were at step 4, so the image creation process
should not try to sync files to nodes (they are not supposed to exists at this
STEP). sync_files sould sync gentent suppported files to the newly created
image. It should use the --image-only parameter IMHO (if I understand well
which is not sure).
I'm currently working on sync_files (I've updated the package to have more user
friendly log (replaced ARRAYS(#1232134] by a human readable content)). Despite
my efforts, sync_files does not seem to sync any file. the oscartst user is not
synced to the nodes (and not to the image as Geoffroy disabled the use of sc3
(replaced by c3) because he found issues with sc3).
So I'm working on understanding how sync_files and sc3 work before I can fixe
those issues. I think that fixing sc3 and sync_files should fix many problems.
Please, can you setup your local epel repo and try again, I think that you
should encounter less issues. I think that the image sould succeed. I see no
reason why it would fail.
Best regards.
Olivier.
--
Olivier LAHAYE
CEA DRT/LIST/DCSI/DIR
________________________________
De : Costel Seitan [csei...@slb.com]
Date d'envoi : jeudi 28 février 2013 12:06
À : oscar-us...@lists.sourceforge.net
Cc: LAHAYE Olivier
Objet : RE: [Oscar-users] RE : RE : RE :OSCAR on CentOS-6: how to
test.(Warning, experimental) WAS:Re: OSCAR on CentOS-6: bootstrap successfull
using online repository.
Hello,
Thank you for your comments DongInn. It is good to hear from you again?
I would also like to add my thanks to those of DongInn
for your hard work Olivier. I am sure you will get all that bugs fixed.
I am new to the Oscar-users list but I can talk a little of my experience with
OSCAR.
I used OSCAR three times to setup/upgrade a cluster
and from what I remember there were always few workarounds to apply in order to
make it work.
However, I always thought it was really valuable as a tool and I always find it
very professional.
I believe there is really a need for a bullet-proof OSCAR version for the
current OS versions (like Ubuntu 12 or CentOS 6)
before making big changes further.
In my opinion SIS is more important when setting up/upgrading the cluster and
less afterwards
(I might have used SIS image once after the initial setup when a node was
broken). However, the other tools included
with OSCAR (C3, Ganglia, SGE/Torque, Sync-files to name just a few) are really
valuable tools to manage
and run the cluster.
So far as concerns me, It won’t bother me if there are only few workarounds to
be applied, I can handle them. I just need to
be able to install my nodes quickly and easily once the image was created and
tests with the
different hardware (nodes) performed.
Coming back to the tests, I updated my packages again with the last ones and
launched a SIS image build.
This night SIS image failed. I am suspecting a timeout on epel repositories. I
am looking for a way to bring the
packages locally and eventually create a repo if possible because it is indeed
time-consuming (somehow strange as I see
the mirror is redirecting requests to different sites correctly).
Other bad news: the SIS image I tried this morning failed at the end too. :-(
The GUI showed a ‘User cancelled building image “centos63” ‘ (NOTE: centos63
is the name I gave to the image myself.)
The last messages I had into the terminal window were:
..
=============================================================================
== [sync_file]
=============================================================================
[INFO] The following files do not exist, we cannot synchronize them, we will
see later if we can use a distro template.
[ /etc/modprobe.conf ]
Array: 1 element(s)
ERROR: Impossible to detect the OS for /var/lib/systemimager/images/centos6 at
/opt/sync_files/bin/sync_files line 319
main::distro_detect('dir', '/var/lib/systemimager/images/centos6')
called at /opt/sync_files/bin/sync_files line 798
ERROR: Impossible to detect the distro (/var/lib/systemimager/images/centos6)
at /opt/sync_files/bin/sync_files line 800.
ERROR: Impossible to execute /opt/sync_files/bin/sync_files -i centos63 --force
--image-only at /var/lib/oscar/packages/sis/api-post-image line 205.
ERROR: /var/lib/oscar/packages/sis/api-post-image exitted badly (2) at
/usr/bin/post_rpm_install line 124
ERROR: Couldn't run post_rpm_nochroot for sis at /usr/bin/post_rpm_install line
125
Looking for /var/lib/oscar/packages/switcher/api-post-image...
Looking for /var/lib/oscar/packages/switcher/post_rpm_nochroot...
Looking for /var/lib/oscar/packages/sync-files/api-post-image...
--> About to run /var/lib/oscar/packages/sync-files/api-post-image for
sync-files
Created templates for image centos63 in /opt/sync_files/templates/image/centos63
Make sure these files do not contain any user accounts.
Only system account IDs should be included!ERROR: There were errors running
post_rpm_install scripts. Please check your logs. at /usr/bin/post_rpm_install
line 131.
Looking for /var/lib/oscar/packages/sync-files/post_rpm_nochroot...
Looking for /var/lib/oscar/packages/torque/api-post-image...
Looking for /var/lib/oscar/packages/torque/post_rpm_nochroot...
Looking for /var/lib/oscar/packages/yume/api-post-image...
Looking for /var/lib/oscar/packages/yume/post_rpm_nochroot...
ERROR: Impossible to execute /usr/bin/post_rpm_install centos63 eth0 --verbose
at /usr/lib/systeminstaller/SystemInstaller/Tk/Image.pm line 805
rsync: failed to connect to 127.0.0.1: Connection refused (111)
rsync error: error in socket IO (code 10) at clientserver.c(124)
[receiver=3.0.6]
ERROR: Impossible to do post binary package install, deleting the image... at
/usr/lib/systeminstaller/SystemInstaller/Tk/Image.pm line 805
rsync: failed to connect to 127.0.0.1: Connection refused (111)
rsync error: error in socket IO (code 10) at clientserver.c(124)
[receiver=3.0.6]
ERROR: Impossible to run postimagebuild at
/usr/lib/systeminstaller/SystemInstaller/Tk/Image.pm line 805
Cleaning up centos63
ERROR: Impossible to create the image at /usr/lib64/perl5/vendor_perl/Tk.pm
line 250
Any ideas?
I suspect the rsyncd daemon. I checked and there is a
/etc/init.d/systemimager-server-rsyncd script but
rsyncd daemon running.
Shall I launch it manually? Can I do something to re-build the image without
downloading again all the packages (the slowest part).
Best Regards,
Costel SEITAN
From: LAHAYE Olivier [mailto:olivier.lah...@cea.fr]
Sent: Wednesday, February 27, 2013 6:43 PM
To: oscar-us...@lists.sourceforge.net
Subject: [Oscar-users] RE : RE : RE :OSCAR on CentOS-6: how to test.(Warning,
experimental) WAS:Re: OSCAR on CentOS-6: bootstrap successfull using online
repository.
Hi DongInn,
Cool to see you again; Thanks for your compliment :-)
I've updated several package since yesterday:
opkg-blcr-1.0.2-1.noarch.rpm
blcr-modules_2.6.32_279.el6.x86_64-0.8.4-2.x86_64.rpm
opkg-munge-1.0.4-1.noarch.rpm
opkg-munge-server-1.0.4-1.noarch.rpm
opkg-munge-client-1.0.4-1.noarch.rpm
opkg-torque-2.1.17-1.noarch.rpm
opkg-torque-server-2.1.17-1.noarch.rpm
opkg-torque-client-2.1.17-1.noarch.rpm
openmpi-switcher-modulefile-1.6.3-2.el6.noarch.rpm
I'm sorry with so much updates, but I'm fighting bugs one by one and I'll GET
ALL OF THEM ;-) (at least the most obvious ones ;-))
More seriously, I understand your point on the deployment solution, but before
I need to fix many issues that are revealed by step 8 in fact. For example,
some services are not enabled and running on nodes and step 7 fails sometimes
without error and some services are left unconfigured. Such things will lead to
many user requests, and since I'm in the code with a clear view, I'd prefer
finish most issues.
(for non crutial issues, I add the issue to my todolist. For example, the
openmpi-siwthc-module-file fails to detect the blcr lib because it checks for
/usr/lib64/libcr.so instead of /usr/lib64/libcr.so.0. I just added .0 the the
check code for the moment and added to my todo list to improve the libcr
detection routine.
I'm updated the systemimager 4.2.0-0.92 rpm which should work better than
nothing. For sure, I'll work on the git repo of systemimager soon to have it
work with more recent kernel and most important have an udev that actually load
the required modules. (the issue is that the current systemimager fails to load
modules because udev needs the full modprobe binary and is incompatible with
the busybox binary. (the git code already have that, unfortunately, udev fails
to compile because it uses system kernel headers instead of systemimager
compiled kernel headers. for example it includes input.h which includes
linux/input.h. IMHO, I'll copy all the systemimcludes in the systemimager build
tree before doing installs and add the -I option that will uses those includes.
it's dirty, but should work.
Best regards.
PS: I've also uploaded a few things in the download directory:
blcr-0.8.4-2.src.rpm
openmpi-switcher-modulefile-1.6.3-2.el6.src.rpm
Olivier.
--
Olivier LAHAYE
CEA DRT/LIST/DCSI/DIR
________________________________
De : DongInn Kim [di...@cs.indiana.edu]
Date d'envoi : mercredi 27 février 2013 16:56
À : oscar-us...@lists.sourceforge.net
Objet : Re: [Oscar-users] RE : RE :OSCAR on CentOS-6: how to test.(Warning,
experimental) WAS:Re: OSCAR on CentOS-6: bootstrap successfull using online
repository.
Hi,
It seems that the new repository and trunk version codes work fine with CentOS
6.3 X86_64. :-) I have not fully gone through the deployment steps because I do
not have the test client nodes but I believe that it would work. I have
actually deployed my image on ubuntu 12.04 a few weeks ago. (I spent several
weeks in making the right initrd image.)
I do not really care about the step 7 and step 8 in oscar_wizard because the
cluster is deployed with my image and I have all the tools to manage them as I
wish.
Usually I spend most times in deploying my oscar image to the client nodes
because there are several issues of configuring the SIS image.
- hard to inject the missing kernel module to the generated initrd image (e.g.,
especially network kernel module is not loaded and a client node can not get
the oscar image after pxeboot).
- the firmware place where a kernel module look at is not correct
- ext4 fs is not supported
- a file with the hexdecimal file name representing the client's IP address is
not automatically generated on /tftpboot/pxelinux.cfg/ as soon as a client node
is imaged. If the hexdecimal file is generated, the client node whose IP
address is related to the hexdecimal would not be re-imaged when it is rebooted.
I know that this is not a pure oscar issue but more like the SIS issues.
If we want to make OSCAR more reliable and useful, I think we'd better spend
more time on the deployment solutions but not too much on the particular OSCAR
packages.
Once we have a solid deployment solutions, we can expand anything as we wish.
Thanks Olivier for his hard work.
BTW, would it work if i update the OSCAR repository only with your opkg-munge?
Regards,
--
- DongInn
On Feb 27, 2013, at 8:04 AM, LAHAYE Olivier
<olivier.lah...@cea.fr<mailto:olivier.lah...@cea.fr>> wrote:
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<mailto:csei...@slb.com>]
Date d'envoi : mercredi 27 février 2013 13:48
À : LAHAYE Olivier
Cc: oscar-us...@lists.sourceforge.net<mailto:oscar-us...@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
------------------------------------------------------------------------------
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-devel mailing list
Oscar-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oscar-devel
