Failure to auto-install 9.3/10.0 using same XML file....
Hi everyone, I've been having one of those days, and I would like to know if anyone else has seen the same problems I have been experiencing... I have a Linux server running 9.2, serving installation sources via http for SuSE 9.2, 9.3, and 10. Using a 9.2 DVD, I can quite happily install a new system over the network using the following (IPs changed to protect the not-so-innocent, 10.20.30.59 is serving the sources via Apache): initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/9.2 autoyast=http://10.20.30.59:/conf/infra.xml However, if I use *EXACTLY* the same xml configuration file but this time using a 9.3 DVD... initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/9.3 autoyast=http://10.20.30.59:/conf/infra.xml it fails (as far as I can make out) with an "out of swap space" error. I of course then tried increasing swap, but received the same error, albeit I think a bit further along than before. Its hard to say for sure because it still fails installing the kernel source rpm, just not at the exact same point. Swap is now twice the size of core memory. I then tried (again using *EXACTLY* the same xml file) to install 10: initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/10.0 autoyast=http://10.20.30.59:/conf/infra.xml It got further, but the system appeared to completely hang about 25 seconds from completing the package install. I was unable to get any response at all. The hardware is a brand-new DL380-G4 with 2Gb RAM. The 9.2 installation source server has not been patched to the absolute latest (but why would this matter - I'm only using it to share the installation media, right?). I should also point out I suffer from the old problem with the network timeout due to spanning tree on the switches (don't think it has any bearing on the issue though). The DVDs are straight out of the box, and are the professional versions. The xml config file is nothing fancy - just sets up the partitioning, defines package selection, and runs a post-install script. I searched the net, and trawled through the archives (oh for a search facility ;-) ), but I couldn't find anything. Could someone please point me in the right direction? Thanks! Colin.
UNIX-admin wrote:
Hi everyone,
I've been having one of those days, and I would like to know if anyone else has seen the same problems I have been experiencing...
I have a Linux server running 9.2, serving installation sources via http for SuSE 9.2, 9.3, and 10.
Using a 9.2 DVD, I can quite happily install a new system over the network using the following (IPs changed to protect the not-so-innocent, 10.20.30.59 is serving the sources via Apache):
initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/9.2 autoyast=http://10.20.30.59:/conf/infra.xml
However, if I use *EXACTLY* the same xml configuration file but this time using a 9.3 DVD...
initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/9.3 autoyast=http://10.20.30.59:/conf/infra.xml
it fails (as far as I can make out) with an "out of swap space" error.
I of course then tried increasing swap, but received the same error, albeit I think a bit further along than before. Its hard to say for sure because it still fails installing the kernel source rpm, just not at the exact same point. Swap is now twice the size of core memory.
I then tried (again using *EXACTLY* the same xml file) to install 10:
initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/10.0 autoyast=http://10.20.30.59:/conf/infra.xml
It got further, but the system appeared to completely hang about 25 seconds from completing the package install. I was unable to get any response at all.
The hardware is a brand-new DL380-G4 with 2Gb RAM. The 9.2 installation source server has not been patched to the absolute latest (but why would this matter - I'm only using it to share the installation media, right?). I should also point out I suffer from the old problem with the network timeout due to spanning tree on the switches (don't think it has any bearing on the issue though).
The DVDs are straight out of the box, and are the professional versions.
The xml config file is nothing fancy - just sets up the partitioning, defines package selection, and runs a post-install script.
I searched the net, and trawled through the archives (oh for a search facility ;-) ), but I couldn't find anything. Could someone please point me in the right direction?
I learned (the hard way) that the autoinst.xml format seems to be the same for versions mentioned, and I guess the format is, but the consequence, i.e. what happens under the hood, is quite different between the versions. I've struggled with this on and off thru the years and I have come to the conclusion that every new version warrants changes to the autoinst.xml files, and I'm never sure about the outcome. I guess some xmllint should give you a hint. In any case, I doubt that you ever going to be able to use the same autoinst.xml files between versions. At any rate, I haven't been able to. D'ya need more? Cheers /LarS
Hi Lars, I can understand that if I was doing something funky, and even so far as the package selections assuming changes in the default behaviour - but for it to trigger such a weird error on 9.3 and hang 10.0? It will be a real pain to have to maintain different config files as well - especially as I'm trying to set up a one-stop server that will be providing a company-wide base build for Linux. Does anyone think the problem could lie in the way it is selecting packages? The problems (so far) appear during this phase of the install. I certainly don't know enough about YaST internals to even guess as to why it would cause me to use up all of swap (unless that is a red-herring). Thanks for the insight though - I'll start thinking about maintaining different config trees..... Col. On Tue, 2005-12-06 at 22:26 +0100, Lars Stavholm wrote:
UNIX-admin wrote:
Hi everyone,
I've been having one of those days, and I would like to know if anyone else has seen the same problems I have been experiencing...
I have a Linux server running 9.2, serving installation sources via http for SuSE 9.2, 9.3, and 10.
Using a 9.2 DVD, I can quite happily install a new system over the network using the following (IPs changed to protect the not-so-innocent, 10.20.30.59 is serving the sources via Apache):
initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/9.2 autoyast=http://10.20.30.59:/conf/infra.xml
However, if I use *EXACTLY* the same xml configuration file but this time using a 9.3 DVD...
initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/9.3 autoyast=http://10.20.30.59:/conf/infra.xml
it fails (as far as I can make out) with an "out of swap space" error.
I of course then tried increasing swap, but received the same error, albeit I think a bit further along than before. Its hard to say for sure because it still fails installing the kernel source rpm, just not at the exact same point. Swap is now twice the size of core memory.
I then tried (again using *EXACTLY* the same xml file) to install 10:
initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/10.0 autoyast=http://10.20.30.59:/conf/infra.xml
It got further, but the system appeared to completely hang about 25 seconds from completing the package install. I was unable to get any response at all.
The hardware is a brand-new DL380-G4 with 2Gb RAM. The 9.2 installation source server has not been patched to the absolute latest (but why would this matter - I'm only using it to share the installation media, right?). I should also point out I suffer from the old problem with the network timeout due to spanning tree on the switches (don't think it has any bearing on the issue though).
The DVDs are straight out of the box, and are the professional versions.
The xml config file is nothing fancy - just sets up the partitioning, defines package selection, and runs a post-install script.
I searched the net, and trawled through the archives (oh for a search facility ;-) ), but I couldn't find anything. Could someone please point me in the right direction?
I learned (the hard way) that the autoinst.xml format seems to be the same for versions mentioned, and I guess the format is, but the consequence, i.e. what happens under the hood, is quite different between the versions. I've struggled with this on and off thru the years and I have come to the conclusion that every new version warrants changes to the autoinst.xml files, and I'm never sure about the outcome. I guess some xmllint should give you a hint. In any case, I doubt that you ever going to be able to use the same autoinst.xml files between versions. At any rate, I haven't been able to.
D'ya need more?
Cheers /LarS
UNIX-admin wrote:
Hi Lars,
I can understand that if I was doing something funky, and even so far as the package selections assuming changes in the default behaviour - but for it to trigger such a weird error on 9.3 and hang 10.0?
You'd better believe it. My experience is that package selection has never been a (real) problem. It's everything else. I maintain several Linux distros (Redhat, Fedora, Ubuntu, SuSE, all with different versions) with specific package selections and added packages, all for network installs, and the package selection part works OK. However, slightest change in partitioning could have strange consequences (much like you described). In your case, there is a change in partitioning (or something else) since you are using what you think is a working autoinst.xml on a new version of the distro. New ballgame altogether. /LarS
It will be a real pain to have to maintain different config files as well - especially as I'm trying to set up a one-stop server that will be providing a company-wide base build for Linux.
Does anyone think the problem could lie in the way it is selecting packages? The problems (so far) appear during this phase of the install. I certainly don't know enough about YaST internals to even guess as to why it would cause me to use up all of swap (unless that is a red-herring).
Thanks for the insight though - I'll start thinking about maintaining different config trees.....
Col.
On Tue, 2005-12-06 at 22:26 +0100, Lars Stavholm wrote:
UNIX-admin wrote:
Hi everyone,
I've been having one of those days, and I would like to know if anyone else has seen the same problems I have been experiencing...
I have a Linux server running 9.2, serving installation sources via http for SuSE 9.2, 9.3, and 10.
Using a 9.2 DVD, I can quite happily install a new system over the network using the following (IPs changed to protect the not-so-innocent, 10.20.30.59 is serving the sources via Apache):
initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/9.2 autoyast=http://10.20.30.59:/conf/infra.xml
However, if I use *EXACTLY* the same xml configuration file but this time using a 9.3 DVD...
initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/9.3 autoyast=http://10.20.30.59:/conf/infra.xml
it fails (as far as I can make out) with an "out of swap space" error.
I of course then tried increasing swap, but received the same error, albeit I think a bit further along than before. Its hard to say for sure because it still fails installing the kernel source rpm, just not at the exact same point. Swap is now twice the size of core memory.
I then tried (again using *EXACTLY* the same xml file) to install 10:
initrd netdevice=eth0 hostip=10.20.30.40 netmask=255.255.255.0 gateway=10.20.30.1 install=http://10.20.30.59:/10.0 autoyast=http://10.20.30.59:/conf/infra.xml
It got further, but the system appeared to completely hang about 25 seconds from completing the package install. I was unable to get any response at all.
The hardware is a brand-new DL380-G4 with 2Gb RAM. The 9.2 installation source server has not been patched to the absolute latest (but why would this matter - I'm only using it to share the installation media, right?). I should also point out I suffer from the old problem with the network timeout due to spanning tree on the switches (don't think it has any bearing on the issue though).
The DVDs are straight out of the box, and are the professional versions.
The xml config file is nothing fancy - just sets up the partitioning, defines package selection, and runs a post-install script.
I searched the net, and trawled through the archives (oh for a search facility ;-) ), but I couldn't find anything. Could someone please point me in the right direction? I learned (the hard way) that the autoinst.xml format seems to be the same for versions mentioned, and I guess the format is, but the consequence, i.e. what happens under the hood, is quite different between the versions. I've struggled with this on and off thru the years and I have come to the conclusion that every new version warrants changes to the autoinst.xml files, and I'm never sure about the outcome. I guess some xmllint should give you a hint. In any case, I doubt that you ever going to be able to use the same autoinst.xml files between versions. At any rate, I haven't been able to.
D'ya need more?
Cheers /LarS
On Tuesday 06 December 2005 22:26, Lars Stavholm wrote:
and I have come to the conclusion that every new version warrants changes to the autoinst.xml files, and I'm never sure about the outcome.
I consider such a behaviour as a bug unless I've documented it that way. If you hit such a case and it's not documented, please inform me (preferred way is bugzilla). -- ciao, Uwe Gansert Uwe Gansert, Server Technologies Team SUSE LINUX Products GmbH, Maxfeldstrasse 5, D-90409 Nuernberg, Germany e-mail: uwe.gansert@suse.de, Tel: +49-(0)911-74053-0, Fax: +49-(0)911-74053-476, Web: http://www.suse.de
It sounds to me as if the 9.3 kernel is only detecting a fraction of your memory. What happens if you apply 'mem=2000M' as a kernel boottime parameter? Any information on that in the boottime output? Bjørn -- Bjørn Tore Sund Phone: (+47) 555-84894 Stupidity is like a System administrator Fax: (+47) 555-89672 fractal; universal and Math. Department Mobile: (+47) 918 68075 infinitely repetitive. University of Bergen VIP: 81724 Support: http://bs.uib.no Contact: teknisk@mi.uib.no Direct: bjornts@mi.uib.no
participants (4)
-
Bjorn Tore Sund -
Lars Stavholm -
UNIX-admin -
Uwe Gansert