On Wednesday 21 April 2004 14:30, Lee Mayes wrote:
I'm using some old hardware (and vmware instances) to verify some SuSE 9.0 autoinstalls. Some only have 128MB of RAM. While I wouldn't attempt to run any production services on these boxes, they're a good test of my final scripts and processes.
I've run across a problem in the way the rpm's are installed using a network (http) source. By pushing a virtual console during the actual install, it appears that the rpm command is caching the rpm to be installed in / (the ramfs) vs. the mounted hardrive. So basically, this limits the size of an rpm that can be installed from a network source to roughly Physical Ram - ( kernel + cramfs installer ram footprint), which works out to roughly 52MB on a 128M system. With less than 128M there are other concerns, so that's my bare minimum for testing installs.
Is there a switch I can pass AutoYaST2 somewhere to have it use the installing system's mounted /var/tmp vs. the ramfs? There are some pretty big RPM's out there, like OpenOffice (who's integration I need to test :^).
Funny you mention this, because I ran into the same limitation, only worse: the kernel rpm was too big! The target PC was a P133 with 96 MB ram (the install process was using a 200 MB swap partition) After a bit peeking around I discovered a solution: ln -s /var/adm/YaST/InstSrcManager/IS_CACHE_0x00000001/MEDIA/i586 /mnt/tmp (The target system is mounted under /mnt [from memory, hope I got that right ;)] ) When I put that in a pre-installation script, the counter measure was simply creating another IS_CACHE_0x00000002. So this had to be done /during/ the installation process. Just find a free text console (<Ctrl>+<Alt>-<F1> or so) or login remotely with a second SSH login. Pretty clumsy, but it worked. Indeed it would be very handy to supply a kind of temporary working space, either on the target partitions (usually there will be some free space left) or by supplying a partition that can be used (possibly after being formatted). Cheers, Leen