| Revision History | ||
|---|---|---|
| Revision $Revision: 1.7 $ | $Date: 2000/08/16 14:41:54 $ | |
Redhat 6.2 uses a different version of RPM from Redhat 5.2. This means that upgrading necessitates a complete reinstall. This is a good idea in any case to hoover out old log files, zap any hidden user files etc.
You need a new boot floppy or boot CD (for portables)
Change the lcfg include line <linux_rh52.h> to <linux_rh62.h>
Ensure that partitions that you wish to be preserved across the install are marked preservein the relevant lcfg line. You shouldn't just rely on this - you should also take a backup of the relevant partition before upgrading.
If the machine has its own rpmcfg file, copy it to the 6.2 rpmcfg directory (/export/local/linux/6.2/rpmcfg)
Use the install object on the relevant Sun install server to remove the machine and then add the machine (to change the bootptab entry to use the 6.2 NFS installroot) This action is not necessary for portables using the boot CD as the installroot is fetched from CD rather than over NFS.
Boot off the boot floppy and install
Boot floppy images now live in /export/local/linux/images/floppy
The boot floppy is no longer called the install floppy as it can now "safely" be used to debug malfunctioning machines.
The standard rh6.2 NFS boot floppy is 6.2/boot_nfs_date. You should always choose the image with the latest date. It is also a good idea to check that you have the latest boot floppy before upgrading/installing a machine - you can do this by comparing the result of running md5sum on the latest floppy image file with your floppy. For example :-
[don]ajs: cd /export/local/linux/images/floppy/6.2 [don]ajs: md5sum /dev/fd0 33d67eaa0da4c3d2d53451beb3f954f2 /dev/fd0 [don]ajs: md5sum boot_cdrom_120600 33d67eaa0da4c3d2d53451beb3f954f2 boot_cdrom_120600 |
CDROM images now live in /export/local/linux/images/cdrom
There is currently only one CD image file for rh6.2 - the boot cdrom for portables, with filename 6.2/bootcd.iso_date.
The boot cdrom contains both the boot floppy (with the kernel) from which to boot off and the contents of the NFS installroot.
You can blow the CDROM on eskin using, eg.
cdrecord -v speed=4 dev=6,0 /export/local/cdrom/6.2/bootcd.iso_120600 |
You should build RPMs for rh6.2 on a rh6.2 machine to ensure that the correct header files are used. You must rpmsubmit RPMs for rh6.2 on a rh6.2 machine as the format of the RPM dependency files is different between the two OS versions.
RPM under rh6.2 deprecates the use of the ~/.rpmrc file in favour of ~/.rpmmacros. See the section "RPM Configuration using Macros" in the text file /usr/doc/rpm-3.0.4/macros
In order to minimize the amount of rubbish sent to the console at boot time, much of the output from the various programs used at install time are now directed to an alternate virtual console ( accessible through CTRL-ALT-F2). You can page back through the info using SHFT-PageUp.
There have been some transient problems with installing on Dell GX1 machines - the machine fails to mount its NFS root even though it appears to have received its bootp response correctly. If this happens please try and contact AJS so he can look at it. Dell GXA machines are fine.
Mobile IP
Various LFCS packages - coq, hugs98, isabelle, lego, smlnj, cml, eml, ProofGeneral, camlp4, ocaml
Pine
Teaching software - VTK, argo
XDesigner
At install time, the pcmcia cardmgr daemon complains about some script being missing. This can safely be ignored.
A [FAILED] message is displayed on laptop boots. This is because the network configuration code is run before the pcmcia startup code - the network code notices that there isn't a card and delays configuration (it expects the pcmcia code to configure the ether card). The rh6.2 network code wrongly reports this as a failure, so this message can safely be ignored.
This install time error message means that you have extended/logical partitions on one of your drives - the update object can't cope with these so you should set update.dopartitions_drive to no.
The resource auth.logindevperms has been replaced by auth.consolepermclasses and auth.consolepermrules. See the documentation for the object and also man console.perms.
It looks like NDBM isn't functioning correctly with perl.
You can safely ignore this message if you see this message at install time and everything otherwise appears fine.
You can safely ignore this message if you see this message at install time and everything otherwise appears fine.
It is essential that "laptopupdate" is run, followed by a reboot, after an install on a laptop. This is because some RPMs are deliberately not installed during the install process.
Note that order of naming of eth0 and eth1 will depend on which kernel you're using. eg in Don, with the server kernel (drivers compiled in) if we have a Pro100 card in slot 1 (counting from cpu) and then add a 3c905 in slot 5, the pro100 becomes eth1 and the 3c905 is eth0. If we reboot with the public kernel, and have conf.modules say that eth0 is a pro100, then the pro100 card will indeed be eth0. This has implications for install time as the install kernel uses the server kernel. Moral of the story is, if you have two ether cards, you should really use the server kernel to avoid confusion.
Both acroread 3 and acroread 4 are installed, as acroread3 and acroread4 respectively. The command acroread points to acroread 4. Netscape 4.7 (standard redhat 6.2) only works with the acroread 4 plugin.
Local TeX and LaTeX are now installed.
Cdrecord no longer needs to be run as root.
The boot wire no longer needs to be connected to eth0.
tkgnats is now configured for local use
The recent problems with bootpc failing during installs should now be fixed. Please report any further problems.
The netscape support files have now been installed so PDF, DVI and PS support should now be working.
rsync was hanging on transfers of large numbers of files. I think an updated version of rsync has fixed this - if you experience any more problems please contact me so that I can attempt to narrow down the circumstances under which it's occuring.