Hi Joe,
On 02/13/09 16:21, Joseph J VLcek wrote: > Jan, > > I've looked over these changes and they look good to me. > > I'll assume that testing has ensured the correct list of items have > been removed. This is my assumption as well :-) Me and Mary did couple of installations using that reduced image. That said, there is always possibility I might overlook something and this is part of the reason I waited with those changes after 2009.1 was released. I think that now we should have enough soak time to give those changes a try :-) Thank you very much for review ! Jan > > Joe > > > > jan damborsky wrote: >> Hi, >> >> could I please ask for reviewing changes for creating >> reduced AI image ? >> >> Those changes are delivered under umbrella of following bug: >> >> 4166 The automated installer hangs on a 1 GB system without swap >> http://defect.opensolaris.org/bz/show_bug.cgi?id=4166 >> >> webrev: >> http://cr.opensolaris.org/~dambi/bug-4166/ >> >> Special thanks to: >> * Dave Miner for making all investigation and original work >> * Mary Ding for huge help with testing >> >> Background: >> ----------- >> Those modifications reduce the content of AI image and take >> advantage of split of SUNWgui-install package (bugs 5468, 5657). >> This package originally delivered both install libraries required by AI >> and GUI installer. Because of this, bunch of unnecessary stuff >> was being pulled into AI image (X, Gnome libraries, media libraries, >> ...). >> By taking advantage of that split, AI images could be reduced by ~45%. >> >> Testing >> ------- >> These changes are delivered under umbrella of bug 4166, since they allow >> to install on x86 machines with 1GB of available physical memory without >> creating swap device. >> >> For x86, I tested on Ultra 20 with 1GB of memory. Before AI >> installation started, 30MB file was created in /tmp directory >> to make sure there is still some reserve available. >> >> As far as Sparc is concerned, I tested on T1000 (8GB RWM) >> and T5220 (1GB RWM + 512MB swap). >> >> Mary did testing on x86 as well as Sparc sun4u & sun4v machines. >> >> These changes are not intended to address problems with the installation >> on x86 machines with less than 1GB available memory and Sparc machines >> with less than 2GB available memory. That said, it was observed >> that AI succeeded in additional following cases: >> >> * installation in Virtual Box with guest assigned with 600MB of memory >> + 512MB swap >> * Sparc machines with 1GB physical memory + 512MB swap >> >> AI failed in following test cases: >> >> * x86 machines with 512MB of memory >> * x86 machines with 768MB of memory (no swap) >> * x86 machines with 1GB memory, but shared with graphic controller >> >> Mary, please feel free to correct me or add additional information. >> >> >> Additional issues >> ----------------- >> Currently, both AI as well as LiveCD share the same set of steps >> as far as creating and configuring SMF repository is concerned - >> they are implemented by mkrepo script. Since some of services >> are not present in reduced AI image, Distribution Constructor >> started to display following warning messages when building >> reduced AI image: >> >> ... >> ==== br-config: Boot root configuration >> Configuring bootroot... >> Preloading SMF repository... >> Loading smf(5) service descriptions: >> Loaded 120 smf(5) service descriptions >> svccfg: Pattern 'application/font/fc-cache' doesn't match any >> instances or services >> svccfg: Pattern 'system/dbus' doesn't match any instances or services >> svccfg: Pattern 'application/opengl/ogl-select' doesn't match any >> instances or services >> svccfg: Pattern 'application/graphical-login/gdm' doesn't match any >> instances or services >> Creating additional symlinks in ramdisk ... >> ==== ai-br-config: Auto Install boot root configuration >> ... >> >> Those doesn't prevent building AI image, but definitely should be >> addressed. >> I was thinking if solving that problem might fall under bug which >> seems to >> take care of redesigning the stuff mkrepo script takes care of right >> now, >> or if new bug should be filed or if different approach might be more >> appropriate >> to deal with this issue. Please let me know, what you think. Thank you. >> >> _______________________________________________ >> caiman-discuss mailing list >> caiman-discuss at opensolaris.org >> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss >
