jrr@lanl.gov (John R. Red-horse) (07/11/90)
Hello Netlanders, I am a novice (and reluctant) system manager for a small Ultrix LAN consisting of four diskless VAXxtation 3100's being served by a VAXstation 3200 boot node and a DECstation 5400. Each of these systems is running version 3.1 Ultrix and the VAXstations run UWS 2.2 as well. I ran into a peculiar, though self-induced problem that I am hopeful can be recovered from in some reasonable fashion. Specifically, while adding disks to each of the machines, I noticed that I was going to have to modify the config file via the utility ``doconfig'' and rebuild my kernel. Silly me, I asked to have the existing file overwritten when queried by doconfig (the question doesn't say *WHEN* the overwriting gets done) and, quite simply, the config files that I had built originally, through the automated build procedure on the install tapes are gone (sigh). Is there a sane way to do this again without going through the install again from scratch? Assuming that I do have to do the install, do I lose all the stuff that is currently configured and running on the systems (e.g., DECnet, nfs, etc.)? Do not mistake the overall tone of this note as that of a calm man---I am desperate, I need help. Thanks to all in advance. Cheers, John Red-Horse INTERnet:JRRedho@sandia.gov Org 1524(Engineering Analysis Div) Sandia National Laboratories Albuquerque, NM 87185-5800 Bell: (505) 846-6115
alan@shodha.dec.com ( Alan's Home for Wayward Notes File.) (07/15/90)
In article <56677@lanl.gov>, jrr@lanl.gov (John R. Red-horse) writes: > Hello Netlanders, > > [ A description of how the user has overwritten his configuration > files on various ULTRIX systems running V3.1 ]. First, do you have backups of the systems? Restoring them from backup is much easier than regenerating them. > Is there a sane way to do this again without going through the install > again from scratch? Doconfig(8) without any options uses the device list in the running kernel to build what it belives is one the system. It won't add DECnet, but that's easy (two lines). If the current system doesn't know about some newly installed device you can arrange for doconfig(8) to know about it. You should have a file called /genvmunix leftover from the installation. This kernel has close to everything built into it. What you want to do is arrange for it be running when you run /etc/doconfig. [ NOTE: I don't have much experience with diskless systems so this procedure may only be suitable for non-diskless systems. ] 1. Save your current kernel somewhere: # cp /vmunix /vmunix.old 2. Arrange for /genvmunix to be the running kernel on reboot: # cp /genvmunix /vmunix # shutdown -r now "Some reason" 3. Now you can run /etc/doconfig and it should find all the supported devices that are on the system. If you added any special device drivers that the generic kernel doesn't know about you'll have to do those over (the driver for the MD300 page scanner for example). 4. Other extras like DECnet, the trace facility that doconfig(8) doesn't know much about you'll also have do by hand. DECnet support consists of two lines: options DECNET pseudo-device decnet 5. Read the Guide to System Configuration File Maintainence for more about what is in the configuration file. > Assuming that I do have to do the install, do I lose > all the stuff that is currently configured and running on the systems > (e.g., DECnet, nfs, etc.)? Do not mistake the overall tone of this > note as that of a calm man---I am desperate, I need help. You should never have to re-install a system just to get back a configuration file. If all else fails you can hack stuff out of the GENERIC configuration file to get back to something useful. I think there also example configuration files for diskless systems laying around somewhere (or in the DMS documentation). If you expect to add other devices over time it might be a good idea to put them in the configuration file now. Except for whole device drivers extra devices don't take up that much extra space. For example on a DECsystem 5400 that has two KDA50s you can as many as eight RA type disks. Go ahead and put entries for all eight: controller uda0 at uba? controller uda1 at uba? controller uq0 at uda0 csr 0172150 vector uqintr controller uq1 at uda1 csr 0160334 vector uqintr # disk ra0 at mscp drive 0 disk ra1 at mscp drive 1 disk ra2 at mscp drive 2 disk ra3 at mscp drive 3 disk ra4 at mscp drive 4 disk ra5 at mscp drive 5 disk ra6 at mscp drive 6 disk ra7 at mscp drive 7 Please note one of the features that I make use of. Ordinarily the configuration file would have one of uq0 or uq1 for the "disk rax at blah". If you use unique unit numbers you can use "mscp" to tell ULTRIX to figure out which disk is on which controller. Since two KDA50s can support eight disks you can setup the kernel to eventually recognize eight disks as long each one has a unique unit number and is one of those listed. Using "wildcards" like this depends on the drivers. Look at the GENERIC configuration file for more examples of how the other devices drivers do it (or don't). In the case of "mscp" for DSA disks (and tapes, but I won't discuss those) as long as the unit numbers are unique you can use this for any DSA disk on the appropriate controller: RDxx on the RQDX{1,2,3} RAxx on the HSC, KDB50, KDA50, UDA50. RFxx on a DSSI adapter or KFQSA. For disks and controllers like the RDxx/RQDX# that don't allow setting unit numbers this gets harder, but you can still do a little. > > Thanks to all in advance. > Cheers, > > > John Red-Horse INTERnet:JRRedho@sandia.gov > Org 1524(Engineering Analysis Div) > Sandia National Laboratories > Albuquerque, NM 87185-5800 Bell: (505) 846-6115 -- Alan Rollow alan@nabeth.enet.dec.com
grr@cbmvax.commodore.com (George Robbins) (07/15/90)
In article <56677@lanl.gov> jrr@lanl.gov (John R. Red-horse) writes: > Hello Netlanders, > I ran into a peculiar, though self-induced problem that I am hopeful > can be recovered from in some reasonable fashion. Specifically, > while adding disks to each of the machines, I noticed that I was going > to have to modify the config file via the utility ``doconfig'' and rebuild > my kernel. See the man entry for doconfig(8). There is no reason you can't run it again and regenerate the configuration files. The trick seem to be that you have to be running "genvmunix", since what it does is scan through all the myriad devices defined in this kernel and echo back those that autoconfig actually found when the kernel was booted. > Is there a sane way to do this again without going through the install > again from scratch? Assuming that I do have to do the install, do I lose > all the stuff that is currently configured and running on the systems > (e.g., DECnet, nfs, etc.)? Do not mistake the overall tone of this > note as that of a calm man---I am desperate, I need help. You can also probably start the installation and abort it part way, but you have to be very careful about this, especially if you don't know/ remember the flow of events. It should alo be mentioned that occasional backups of *all* the disks, including the "system areas", help prevent this sort of embarressment. -- George Robbins - now working for, uucp: {uunet|pyramid|rutgers}!cbmvax!grr but no way officially representing: domain: grr@cbmvax.commodore.com Commodore, Engineering Department phone: 215-431-9349 (only by moonlite)