Hello, Sorry for the later summary, but of course I wanted to finish the upgrade before writing. Thanks to the following (in no particular order) for responding: John Elser <jElser@ck8.uscourts.gov>, Tim Evans <tkevans@tkevans.com>, Kevin Buterbaugh <Kevin.Buterbaugh@lifeway.com>, Kevin Reichhart <KReichhart@yantra.com>, David Foster <foster@dim.ucsd.edu>, "McCaffity, Ray (Contractor)" <McCaffityR@epg.lewis.army.mil> Basically everyone said that they would definitely upgrade vs. install from scratch - This is what we did and everything came out fine! David Foster also included some helpful notes as well as a Sun blueprint document (see the end of this email). First we coppied the disk with our OS file systems (/, /var, Etc) to a "new" disk that we weren't using for anything, so that we could revert back to the old disk in the event something unexpected happened. At the suggestion of a few list members I had looked into live upgrade, but wanted to understand the inner workings of live upgrade a bit more through experience before trying it out on our primary server. We coppied the partitions from the existing disk to the new disk by dumping it to a file with ufsdump, and then restoring that file to mounted partitions on the new disk. We originally tried using ufsdump piped directly into ufsrestore, but got lots of error messages about not being able to restore files because they were not on the volume. We next installed boot blocks on the new drive with: installboot /usr/platform/`uname -i`/lib/fs/ufs/bootblk /dev/rdsk/c0t12d0s0 Once we switched disks around so that the copy of our OS disk was c0t0, and successfully booted from the "new" disk, we made sure that the latest recommended patch set was installed and then commented out all unnecessary partitions in /etc/vfstab (home partitions, /export). Before starting the upgrade we deleted /etc/init.d/sshd, /etc/rc3.d/S75sshd, /etc/ssh/, and /etc/ssh2/ because we wil be using OpenSSH now instead of the comercial SSH client and server we have been using. After booting from the Install CD (apparently you can't boot from the DVD), letting the installer copy itself to our swap partition, and then using the DVD to perform the upgrade (which took about two and a half hours) we rebooted and fixed a few minor things which the upgrade had broken: We restored our own /etc/init.d startup script for samba (it had been replaced with a script for samba which gets installed in /usr/sfw), restored our slightly C2 security tweaked copy of the Makefile in /var/yp (it had been overwritten with a new Solaris 9 version which we will soon properly modify as we did the Solaris 8 Makefile, and recreated our sendmail.cf from the .mc file which we luckily still had (the old sendmail.cf file did not work with sendmail 8.12.2) so that VirtUserTables (per-domain email aliases) and GenericsTable (rewriting of outgoing email addresses) kept working. Here are the notes that David forwarded along, as well as my original post below. Thanks again to all who encouraged that I not reinstall. ---------- Blueprint document provided by David Foster: http:// www.sun.com/blueprints Part No.: 806-5333-10 Revision 01, April 2000 ---------- SOLARIS_UPGRADE_CHECKLIST 11-29-01 DSFoster -------------------------------------------------------------------------------- Upgrade Troubleshooting Tips Before starting an upgrade, please check for these common problems. I. BACKUP! Do a level 0 dump of all filesystems before doing anything under this document. II. /etc/vfstab - Comment out all NFS entries. - Check for multiple swap entries. Comment out all except for one simple disk based swap partition. - Comment out all data (or non-O.S.) partitions. - Make sure root partition is on slice 0 - Having entire O.S. in slice 0 is OK. - Make sure there are no /dev/md/ devices. If any exist, see DISKSUITE section below - Make sure there are no /dev/vx/ devices. If any exist, see Volume Manager section below III. FSCK - Make sure devices to fsck are 2 GB or smaller. - Boot from cdrom and run fsck by hand to fix problems prior to upgrade. IV. LINKS - Watch out for links that use absolute paths. Upgrade will overwrite links that it can't reference. This could cause the partition to run out of disk space. - Remove packages that will make changes to automounted directories. - If symbolic links exist from /var/sadm, either restore the files on the /var partition or do a full install. V. /var - Make sure /var is under root or a separate partition. /var should not be a link. - /var/sadm must not be a link or a separate partition. Must be directly under /var. - If symbolic links exist from /var/sadm, either restore the files on the /var partition or do a full install. - Run upgrade.chk to make sure /var/sadm/install/contents is not corrupt. - Make sure /var/sadm/softinfo/INST_RELEASE exists and the contents look like; OS=Solaris VERSION=2.3 REV=0 ------------------------------------------------------------------------------ DISKSUITE DISKSUITE needs to be disabled on OS disks prior to the OS upgrade. The actual upgrade procedures for DISKSUITE-systems are found in different places depending on the version of DiskSuite. You might want to check Appendix D of the Solstice Disksuite 3.0 through 4.1 Administration Guide. For DiskSuite 4.2 and above, there is a seperate installation and upgrade document. Make certain that the new version of Solaris that you are upgrading to can support the version of DiskSuite you are running, otherwise it may be necessary for you to upgrade DiskSuite as well. ------------------------------------------------------------------------------ VOLUME MANAGER Volume Manager must be disabled on OS disks prior to upgrade of the OS. Refer to the Volume Manager Installation Manual for complete procedures for upgrading Solaris with Volume Manager. Even if you are not upgrading Volume Manager, it is necessary to perform some upgrade functions with Volume Manager - these steps are documented in the VxVM Installation Guide. ---------- > > Hi all, > > I'll be faced with this project in the coming months and wanted to see > > what you all think. We have an E250 running Solaris 8, along with various > > SunFreeware packages and software we compiled ourselves (all in > > /usr/local). I am pondering upgrading the server from Solaris 8 to > > Solaris 9 vs. installing Solaris 9 from scratch - I have never been much > > of an upgrade guy, I prefer to keep my data partitions (home, mail spool, > > Etc) and instal from scratch wherever possible. My concern in doing this > > however, is that I would still like all of my 3rd party software in > > /usr/local to work once I upgrade. The compiled stuff will probably work > > fine (depending on any configuration files which may live in /etc), > > however I would like to retain the ability to uninstall or upgrade one or > > more SunFreeware packages at a later date. IF I install Solaris 9 from > > scratch, there goes my package database correct? I can't scheme a way to > > keep package info for the SunFreeware packages so I can later interject > > them into a new package database (after installing Solaris 9). I can't > > just copy over the package database from Solaris 8 for obvious reasons (it > > will hose the package database from the Solaris 9 install). > > > > What are folks typically doing in this kind of situation? I look > > forward to your responces, and of course will summarize. > > > > Thanks, > > Ivan Fetch. _______________________________________________ sunmanagers mailing list sunmanagers@sunmanagers.org http://www.sunmanagers.org/mailman/listinfo/sunmanagersReceived on Mon Aug 12 15:00:54 2002
This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:42:51 EST