rpruess@umaxc.weeg.uiowa.edu (Rex Pruess) (06/13/90)
A few sites have posted notes about their experiences in upgrading to UMAX 4.3 BSD. We found the notes useful & hope that you find our notes useful too. General Comments We upgraded two Encore Multimax 320 systems to UMAX 4.3 BSD R4.0.0. One of our Encores was running UMAX 4.2 BSD R3.3.0 & the other was running UMAX 4.2 BSD R3.1.2. The upgrade process was very time consuming. We began the upgrade at 4 a.m. Saturday and finished at 6 p.m. Sunday. We also moved some hardware between the two systems for performance reasons. Yes, we slept a few hours too. Nonetheless, reserve plenty of time for the upgrade. It is wonderful to be running UMAX 4.3 BSD. No halts, no hangs, few user complaints, so far... Thanks, Encore. Just wish it would have been available earlier. It is difficult to write upgrade notes to cover all possible problems. These comments are not intended to reflect poorly on Encore. We are happy with our Encores. We were able to complete the upgrade without making a call to Encore for assistance. Although the upgrade notes aren't perfect, we were able to complete the upgrade. Installation Comments Step 1 of the installation notes says to be sure to be in multi-user mode. Step 4 says to backup all of your file systems. The backup should be done in single user mode unless you don't care about losing mail, etc. One of our Encores could not read either 4.3 installation tape. We noticed in the "show configuration" listings that three of our four EMCs displayed as EMCNCR and the fourth EMC displayed as EMCDTR. We replaced the odd EMC with our on site spare and then we were able to read the tape. The 4.3 restore command would not read our R3.1.2 backup tapes. We had to use the R3.1.2 restore command to read the tapes. The installation notes states they were written under the assumption that you are upgrading from R3.3.x. Except for this problem, we did not have any other problems specific to upgrading from a pre-R3.3.x system. Carefully read the comments in step 11 of the installation notes. Much of that step is not necessary unless your primary disk is not '0 4 0 16.' This section of notes should be italicized or highlighted in some way. We, too, found it necessary re-partition our disks after the automated script completed. This required another dump/restore of /usr & /var. The automated script makes the /usrx partitions way too small, leaving most of the disk unused. Be sure to verify partition sizes. Even so, we liked the automated script in that it did not burden you with a million questions. The automated script tells you to put the system in auto boot mode. We elected not to do this since we did not want to lose mail, have users login, etc. When you get to this part of the installation, proceed carefully. As mentioned in earlier notes, many devices show (show command) up as being bootable. The default boot device was bkuproot. Since we were prepared for this, we used devconfig on the hardcopy console to fix this problem. devconfig is not my friend either. Be sure you are familiar with devconfig in advance. We created two bootable devices: root (0 4 0 16 0) & mt0 (0 4 0 8). Be sure to select the boot order as '1' & '2' respectively. Don't take the default of '0'. This took about seven pages of console paper. Perhaps we can have a contest to see who can do it in the fewest pages. :-) The mkfs command requires the size parameter. The man page indicates it is optional. mkfs should calculate the size by itself. In the /etc/rc file, the ifconfig netmask was 255.255.254.0. A value of 255.255.0.0 is a more reasonable default. A prompt would have been nice. Because we had two EMCs, an unwanted second hostname of en05 was in this file. We edited the file appropriately. The release notes (page 5) discuss the new groups number for bin (10), staff (20), and other (30). Prior to the upgrade, we changed all of our files to reflect these group numbers. After the upgrade, we noticed bin was still in group 3 in the /etc/passwd file. Hmmm. The installation notes do not address Encore layered products (e.g., ada, cdb, epf, f77, lisp, pascal.) Except for f77 & pascal, we merely copied the relevant files from our backup tapes. Finding them was not trivial. We installed f77 & pascal from the original release tapes. f77 installed okay. The pascal Makefile had to be modified to include a null else in the "is_su:" code. We added the line 'else echo "Hi";\' to correct the problem. See the release notes (page 6, new shell) for details about make problems. The installation notes do not address third party software package restoration. We restored them from backup tapes. If you don't know where all of the files reside, you're in for some fun. If you are running with disk quotas, be sure to read page 5 of the release notes. You need to run the quotconv utility. Production Notes The 4.3 quota command only displays information if there is a problem with your quota. To get the old style output, use the "-v" option. Also, be sure users have read rights to /etc/mtab or the quota command will only display the header line without the actual quota values. The man tree has been substantially reorganized. Take a look at section 3. -- Rex Pruess, Weeg Computing Center, University of Iowa, Iowa City, Iowa 52242 rpruess@umaxc.weeg.uiowa.edu | rpruespb@uiamvs.bitnet | 319-335-5452
loverso@Xylogics.COM (John Robert LoVerso) (06/13/90)
> In the /etc/rc file, the ifconfig netmask was 255.255.254.0. A value of > 255.255.0.0 is a more reasonable default. That's funny ... 255.255.254.0 is what Encore uses on its inhouse network! John