Title: Re: [Oscar-devel] SUSE Linux 10.0 support status
Hi Erich:
 
Image creation under SUSE Linux 10.0 is now working.  Do we need to worry about the directories ownership?
 
BTW, I noticed that the image creation progress bar currently loops - i.e. it goes through a first phase quickly (full progress), then it starts over again and then it fully completes.  I think this is a side-effect of bootstraping?  Anyways, this is a minor issue I guess.
 
I have checked in all my other code modifications.  Still haven't checked in TORQUE and Maui yet - will do that real soon...  (didn't want to affect Fedora Core 4 testing...)
 
There is still one minor issue - currently some of the scripts expect the C3 binaries (cexec, cpush) to be in the PATH.  However, this does not seem to be the case for SUSE and thus before you run "Complete Cluster Setup", you need to quit the wizard, do something like "ssh localhost" to get the new environment variables, re-run the Wizard then "Complete Cluster Setup" will work properly.
 
Do you have any suggestions on how to get around this?
 
Other than that, we need to look into the parallel libraries issue...
 
I'll now take a look to see why "pbsnodes -a" is giving me gibberish... :-)
 
Cheers,
 
Bernard


From: Erich Focht [mailto:[EMAIL PROTECTED]
Sent: Tue 20/06/2006 06:08
To: Bernard Li
Cc: [email protected]; [EMAIL PROTECTED]
Subject: Re: [Oscar-devel] SUSE Linux 10.0 support status

On Tuesday 20 June 2006 10:15, Bernard Li wrote:
> Update...

> It works! (well sort of...)

> Currently there are 2 issues with SystemInstaller and SUSE Linux:

> 1) directory permissions - the scripts tried to chown <user> but <user> does not exist, and thus using root
> 2) services not started on boot, chkconfig --list on the chroot image almost empty

> 1) is not a big issue right now, I'm sure we can somehow fix it.

> 2) is the bigger issue - right now I got around this by manually enabling the services in the chroot, here's a list I came up with:
>
> [execute in chroot, i.e. chroot /var/lib/systemimager/images/<imagename>]
> chkconfig boot.* (need to resolve dependencies)
> chkconfig network on
> chkconfig syslog on
> chkconfig cron on
> chkconfig portmap on
> chkconfig nfsboot on
> chkconfig nfs on
>
> chkconfig ntp on
> chkconfig gmond on (depending if you have Ganglia installed)
> chkconfig pbs_mom on (depending if you have TORQUE installed)
> chkconfig sshd on

Okay, I think I found a way around this mess. It looks like the postinstall
scriptlets are deliberately running differently when started from YAST, this
is not particularly nice.

The solution is:
- prepare the image directory (with /etc/{mtab,passwd,group} inside)
- install (--force --nodeps) glibc into the image with rpm
- install the rest of the stuff with yume

No more permission problems, most of the services are enabled correctly (sshd,
nfs, portmap, ...). Only "network" seems to stay disabled, I'm investigating
this one, still.

Now I need to find a place to add this suse-image-bootstrap stuff. Either
do it similar to "rapt" and add this to "yume" (with special SuSE handling) or
hide it into systeminstaller. I somehow tend towards yume, because
systeminstaller doesn't need to know about the particular distro (since it
only uses packman).

Regards,
Erich

Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Oscar-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/oscar-devel

Reply via email to