SUMMARY: Upgrading several machines to a new SunOS version

From: Henrik Schmiediche (henrik@PICARD.TAMU.EDU)
Date: Sat Oct 16 1993 - 03:44:39 CDT


Thanks to all who responded to my query concerning how to update a
multiple number of workstations to a newer OS. I have not actually
done this myself (yet), but I decided I better post a summary. Anyway,
there are a few interesting suggestions below and I hope this helps
anyone who is facing the same difficulty of upgrading several
machines.

    - Henrik

: I have a cdrom, a SunOS 4.1.3 CD distribution disk and several Sparc
: 1, 1+ & IPX 's that I need to upgrade to 4.1.3 from 4.1 and 4.1.2
: respectively (our Sparc 10's already run on 4.1.3). Any suggestions as
: to what is the easiet way of doing this? One person I know suggests
: setting up one Sparc and then physically moving the hard disk from
: system to system, installing it and doing a disk copy. Anyway, any
: suggestions that could make this repetetive task easier would be
: welcome.

: A second question I have is, outside of security patches, are there
: any strongly recommned Sun OS patches to install. These machine will
: all be using NIS and NFS (clients).

: Any ideas are welcome.

---
>From miker@il.us.swissbank.com Fri Oct  8 03:39:27 1993

The idea of a disk copy (dd of the "c" partition) is good. For the 4.1.2 machines, you could just use sunupgrade, too. The choice depends heavily on how much "non-standard" stuff you have in / and /usr.

--- >From cross@engfs.med.ge.com Fri Oct 8 05:56:42 1993

Oh I know the problem, my solution is a diskless client on a system with cdrom etc and I boot the client from the sever, have a script which formats and partitions the local disk and copies the miniroot into the swap partition and then boot the swap partition, ALL my systems are data less so it only takes a couple of minutes to give the details.

This may sound hard work compared with copying the disks, but you will have to boot each and play around with all the node specified files, doing a sun install does it all for you.

>From poffen@k2.san-jose.ate.slb.com Fri Oct 8 09:58:32 1993

I am doing mine using sunupgrade (available on the CDROM) by mounting the CDROM via NFS on the client, then shutting down to single user (kill 1).

Because the CDROM is kind slow, and I wanted to customize it a bit, I offloaded the CDROM onto disk, and made that available for NFS, and am working on setting it up that way, but straight from CDROM will work.

Note, I am not sure you can use sunupgrade for 4.1 systems, just 4.1.1 and later. You would have to do a full install, but you can still use NFS by loading the miniroot via NFS, this is documented in the Installing SunOS manual.

--- >From mhill@lesol1.dseg.ti.com Fri Oct 8 10:45:20 1993

I would personally use sunupgrade. You can put the CD on one machine and do the upgrades across the net. The entire process should take about an hour for each machine. This includes modifying the etc files and everything.

--- >From tpm@anat.umsmed.edu Fri Oct 8 10:46:03 1993

It took me about 3 hours to backup my system, 1 hour to upgrade my SS1+ from 4.1.1 to 4.1.3 using SunInstall from the Solaris 1.1 CD-ROM disk (into working condition). It took me about 5 hours to chase down some minor problems, but then that is because I am not an experienced SysAdmin. So, I would recommend using the SunInstall package.

I installed patches 100075-11, 100444-42, 100451-55,100452-10, 100512-03, 100529-01, 100573-04, 100988-01, 101080-01, dns-without-nis, and resolv+2.1. The patches were selected from the recommedations from Sun (in their upgrade manual) and from sun-managers summary from a few months ago.

--- >From hsafai@esri.com Fri Oct 8 16:28:04 1993

The easiest method would be to run the upgrade from 4.1.1 to 4.1.3 script. The machine needs to be in single user mode when upgrading. The best to do many machines would be to mount the cd and export it and then nfs mount the cd to the clients that will be upgraded. We upgraded about 15 machines in less than 5 hours.

The best part is that all your existing information will be preserved,i.e., hosts,fstab,exports,mail,printcap, other.

Steps necessary to do the upgrade: 1. / needs at least 2MB free. /usr needs 9MB free 2. Mount te Sun OS media: mount -rt hsfs /dev/sr0 /usr/etc/install/tar

3. Export /usr/etc/install/tar 4. nfs mount that partition on the machine that needs to be upgraded. Make sue the client is in single user mode 5. cd /usr/etc/install/tar/sunupgrade/shell 6. sunupgrade -d -v

You will find the necessary steps on the SunOS CD under /usr/etc/install/tar/sunupgrade.

--- >From warrane.connect.com.au!ups!uniq.com.au!glenn Sat Oct 9 19:04:29 1993

This is a reasonable approach, or you can use dump and restore to copy the stuff over. Just remember to run installboot(8) on the new root filesystem.

--- >From wilcox@kpw104.rh.psu.edu Sun Oct 10 17:21:24 1993

I don't know if you have heard this idea yet, but here goes. We upgraded our system recently by the following way.

Making a prototype server and client by physically making a simple subnet and installing the software and configuring as needed. Then we had every thing that was necessarily installed and did a dump to our /usr/local partition held on our master NIS server. The dumps consisted of / and /usr. (note: before we did this we saved all the necessary files in /etc in a directory for each machine in /usr/local. With this in place, we then booted the servers diskless of the Master server. With the disk free, we could then repartition as needed, newfs and such. Then we did a restore for the files in /usr/local after mounting the disks. ie, mount /dev/sd0a /mnt cd mnt restore if(I think) master:/usr/local/adm/proto.dump.root then the same for /usr After that make sure that resolv.conf is correct, and hostname boot -s install the routesmake sure static routes are up "sh /etc/rc.route" then install the /etc/ files. I think that about covers it... It will save toting disks and tearing apart machines, but might be a little confusing. Some steps might have been missed but I think this might give you an idea. If you need a step by step, I am working on another upgrade, send me a note and I will send you a list of exactly what I did.

--- >From birger@vest.sdata.no Tue Oct 12 09:33:06 1993

I am facing the same problem.

If the customer had been ready for Solaris 2.x, I would have recommended using the local disk for swap, /tmp (tmpfs) and caching, as Solaris 2.3 is supposed to have a lot of goodies coming concerning NFS caching. With a lot of cache, and /usr mounted read-only, diskless clients shouldn't be too much of a burden on the FDDI backbone....

But it seems we have to go for 4.1.3 first. We may convert some of them to swapfull clients just to see how it affects the network.

Please share any info and tools. I don't quite like the idea of having to walk around upgrading ~200 workstations. Even with a pre-configured disk. If I could somehow set up a diskless client that automagically repartitioned it's internal drive when it booted, copied the operating system from the server to local disk, set EEPROM values, and then rebooted from internal disk again.... I would only have to log into the clients and reboot them from the network. -- Henrik Schmiediche, Dept. of Statistics, Texas A&M, College Station, TX 77843 E-mail: henrik@stat.tamu.edu | Tel: (409) 845-9447 | Fax: (409) 845-3144



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:08:25 CDT