On 12/31/09 06:43 PM, Robert Hanke wrote: > Dear Frank: > > > Thanks very much for replying. If you don't mind, I have not yet > properly researched the issue but here are three logical questions: > > > 1. the ok boot -net command and the EEPROM boot-file goes back to my > Sun OS 4.1.3 days. How is this fundamentally different than wanboot? > One ref is: man boot(1M). Can boot -net be used with some kludge or > the other? >
A "boot net" at the OBP invokes the first-stage loader in the OBP, which uses RARP or DHCP to acquire an IP address, and then downloads the second-stage loader (inetboot and wanboot are the two such loaders in existence) that is then used to further load the kernel. So, a network boot always uses "boot net", but in the case of OpenSolaris we only use the wanboot loader, which is downloaded over http and expects to have certain properties available in the OBP. Older OBP's don't have the http support or those properties. > > 2. How difficult would it be (of course $$s too) to issue a minor > sun4U OBP update that addresses only this specific install issue? As > I recall, Sun Service used to call these one-off, untested, UAYOR > modifications "specials". Of course this incurred the customer a fee > at $260 hr, (then) two hour minimum. > It's non-trivial, but I will point out that the most current patches for OBP for platforms that have not reached Sun's end-of-service-life date include the wanboot property support. Older platforms may lack sufficient memory in the OBP itself to include the additional functionality required, so it's not just a matter of porting code. > Given the historical existence of boot -net, I do not intuit that > this issue is as complex as, for example, a driver issue. Rather, it > seems to be an issue of additional command arguments, or file > location and identification. Most of the FORTH code may already exist > from later processor family OBPs. I have seen procedures for OBP > manual editing (again UAYOR!). Are both the old and current OBP FORTH > code sets available for comparison? > OBP code is not open. > > 3. Re the SPARC bootable ISO, I have wondered for a decade just why > is it that an ISO must be burned at all? > In some cases you don't have to burn the ISO to media - many modern firmware environments such as ILOM can emulate a CD-ROM using an ISO file available on the network. However, the real underlying requirement that the ISO expresses is to have a known environment for the installer to execute in. Allowing version skew between the operating system that the installer is run in vs. the system that's being installed creates all sorts of potential problems for development (what API's can I use? what version do we build on?) and vastly expands the testing matrix. If the differences are small (say, build 129 vs. build 130) it would often "just work", but making that true beyond a matter of chance costs real money. > Several years ago I cribbed the good LOFS-based (loopback virtual > file system) CD-ROM ISO un-ISO script at BigAdmin. After some trial > and much error, I extensively modified it into a brute-force series > of progressive subscripts and directories to install Solaris 9 and > later Solaris 10, on a stand-alone host, without a CD-ROM, DVD-ROM, > or a network host. It worked fine for this tar-ball head. > > The biggest issue seemed to be to extract the boot block. But again, > why waste the plastic to burn an OS CD or DVD that may be obsolete in > a week? That is one of the good things about the OpenSolaris > community, rapid updates. Solaris 10 already offers the ability to > load a subsequent OS on a separate partition via Live Update. Is not > there someway to make down-loadable, plastic-less install easier for > the non-enterprise admin or stand-alone user? I also suppose the > single partition ZFS might preclude the latter install approach. > > For background, see: > > [The original BigAdmin article (Singapore? author) is AWOL, and Sun > search appears obtuse.] > > also: > > http://www.sun.com/bigadmin/content/submitted/boot_usb_flash.jsp "I > really wish this were a supported installation method." > > http://wikis.sun.com/display/BigAdmin/Creating+a+bootable+ISO+image > > Installation - How to create ISO images for Solaris 10 U6 (10/08) > SPARC CD 1 and DVD > http://forums.sun.com/thread.jspa?threadID=5349350&tstart=0 > > "Please check previous postings in the various forums. This type of > question has been asked many times. Obviously the image can be > downloaded, burned and booted, since many people have done so. Other > forum replies include questions on DVD media quality, hardware errors > if you try to mount the disc on a running system, etc. Maybe you want > to try the latest release S10 U8 (10/09) and see if you have any > issues." > > > 4. In sum, from the above comment, ironically the reverse issue of > creating ISOs, it seems that there has been regular and reoccurring > discussion of the overall issue of installation via flexible means. > (This is not an identical issue to JumpStart, flash images, automated > remote install to client clusters, or automatic provision of LDOM and > virtual clients, etc.) > > Perhaps a top-level: "How and why do we do this task in this way?" > would be a useful and productive inquiry. Of course, after two > decades of fat-thumbing Sun OS, I might just be approaching that > confused state ... > I'm not sure what you're really asking for here, but I'd suggest starting with the documentation links from www.opensolaris.com. There will be more features in the OpenSolaris installer (indeed, build 130 brought several at once) as we go along; we are not yet at the point where it meets the requirements for the next Solaris release. Dave
