From: Cooper, Tom [tom.cooper] Sent: Monday, June 07, 1999 9:12 AM To: 'sparc-list' Subject: FW: upgrade of 5.1 to 6.0 I've got a Sun ELC (Sun4c arch) with RH 5.1 installed. Over the weekend I attempted to do an upgrade of that machine, and failed. When I ran the upgrade, I went through the process of selecting packages, and cleared the dialog that told me an installation log would be written to the /tmp directory. After that, nothing happened. I waited a couple of hours, I tried again, I tried an install via ftp. Nothing worked. All failed with the same behavior. Has anyone seen this? Regards, Tom This message reflects the personal views and opinions of the author only, and not necessarily those of any organizaion to which the author is related. --
Please provide us with any messages that are appearing on the virutal consoles that relate to this problem.
it appears, based on sparc-list traffic that this is an issue that affects others, as well. Apparently the boxen hang while building the package list because they seem to run out of memory. This is a problem on ELC, and IPX platforms. Is someone addressing this?
It's been quite some time since I last attempted this. I don't recall _any_ messages on the VCs. I'll try to re-run this upgrade over the weekend, and report my findings. R, Tom
Loading 6.0 onto my IPX with 32Mb ran successfully. Of course, when increasing it to 64Mb, it still correctly loads. Either way, the verification of package dependancies does take a very long time.
I retried the upgrade, and found lots of output on the VC's (of course.) How can I capture that data to/from a file? My mouse doesn't seem to be working during the upgrade, so I don't know how to cut/paste info. Any ideas? Is there something I could look for? Please don't tell me that I need to key in all the screen data from VC2 and VC3..... Thanks for the help... ------- Additional Comments From 09/24/99 19:04 ------- I have the same problem on SparcStation 1. The last message in VC5 is: Writing inode tables: done Writing superblocks and filesystem accounting information: done ------- Additional Comments From 09/25/99 14:57 ------- From: Igor Lemm at i.lemm (the Netherlands) Tried with and without formating on Sparcstion 1 (28 Mb): same result. In VC 1 after about 1 hour: install exited abnormally - received signal 10 sending termination signals: done sending kill signals: done unmounting file systems ... /tmp/rhimage /mnt/proc /mnt /proc /tmp you may safely reboot your system ------- Additional Comments From 10/09/99 12:38 ------- From: Igor Lemm at i.lemm And now for the solution .... Make a /tmp/install.log before you hit the last question-screen (might not be important after all). The installation scripts indicates to use this file, yet writes in /mnt/tmp/install.log. Do not install any packages and do not choose for manual selection. Only the bare minimum. At my SparcStation 1, ROM 1.0, 28 MB this started to install RedHat 6.0. Herafter you have to manually install aditional packages, using RPM. Try kdebase* for a start and recursively work your way through all the needed packages. This is not simple because dependecies are stated as libs or files and not as packages. Look at the names carefully ! At this moment I have got RedHat 6.0 and even KDE up and running ! Of course is very slooooowwwww. ------- Additional Comments From 10/25/99 14:33 ------- I have the same problem with new installation of RH 6.0/SPARC via FTP on IPC with 32 MB of RAM. The installation hangs when I try to choose something more than the default set of packages. It would be great to get complete solution to this problem. Regards Piotr Kowalski pkowalsk.pl ----
The crash which is reported in this bug posting (the signal 10) is a result of a bad memory access, probably as a result of running out of memory on the machine. As for the comment about creating a /tmp/install.log file before starting the actual package installation, this is totally and completly wrong. The message says that the installer will write the log to "/tmp/install.log" because after the installation, when someone would want to look at this log file, that is indeed where it is located. However, during the actual installation, the root filesystem is actually mounted at "/mnt/" so the installer actually does write to "/mnt/tmp/install.log" as that corresponds to "/tmp/install.log" after reboot. This bug is getting closed out due to not being a bug, just a memory limitation on the machine.
As the original reporter of this bug, I am not happy that it was simply declared "not a bug." My ELC still does not run 6.x because of this problem. Additionally, I never got any feedback about how to report more information to the developers. Please reopen this bug report, and help me deliver the right information to development so that my problem can be fixed. Thanks! Regards, Tom Cooper
If you are still having problem on your machine and it is not related to memory problems (that is, you have at least 16M of memory for a CDROM/NFS installation and 32M for a FTP/HTTP/Hard Drive installation) then please submit any messages that you are getting on the virtual consoles. I am able to install 6.1 on the ELC that we have in the lab without fail, so please provide any information that you can to help replicate this.
Well, it's been a while since I've futzed with it. My ELC has 64MB RAM, and was running 5.1 fine when I attempted the upgrade. The hard diisk failed, and I have replaced it, but have not gotten RH6x to install properly. I do recall that I've tried 6.0 and 6.1 without success, and that I requested a way to capture electronically the information on the VCs (see the bug notes for my message asking for help with that.) As it stands, I'm sure that it will fail if I try it again. Please tell me how to capture the VC info. I have tried ftp, http, and hard disk installations unsuccessfully. Please advise. Regards, Tom Cooper