From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.5) Gecko/20050109 Fedora/1.7.5-3 Description of problem: When attempting an installation on Feb 05, 2005, the following problems were encountered. could not see lvm installs in text installation, could see the installation using the graphical installer. Inode creation during filesystem formatting seemed very slow and almost stalled. The process was slow for all partitions, but /dev/sda2 that was designated as the home partition was the longest in lagging. All that was left after the installation was a very small directory structure. The directory structure and files created were smalle enough to gzip. I will submit the whole filesystem stub since the residue creation was so small in size. There was a lost-found directory within boot, but no grub subdirectory or kernels. The zip file shows the resulting system. Version-Release number of selected component (if applicable): anaconda-10.2.0.16-1 How reproducible: Always Steps to Reproduce: 1. pick FTP site to install rawhide from 2. attempt installation on several computers. 3. Experience long delays during formatting of filesystem and then installer failure during the installation. Actual Results: A stub filesystem structure that is small enough to gzip and email. Expected Results: Installation of a complete filesystem and no failures with the GUI or during program installation. A working system with no bugs Additional info: The installation was tried from the same mirror and on two different computer types. Similar lagging of filesystem creation was exhibited on both attempts. EXT3 filesystems and swap were the filesystem types attempted.
Created attachment 110698 [details] complete directory structure from faled installation This was so small and included fstab, modprobe.conf and other files in /etc and around the systen. This is why the filesystem was zipped for the information.
What were the symptoms of the failure? Traceback about not being able to mount some filesystems? If so, should be fixed in CVS (and current rawhide), but current trees are almost certainly broken in new and different ways ;)
When I tried to fresh install, I was able to make it through selecting packages and setting up the desired partitions for the new installation. sda1 was /boot and was to be 101 mb, sda2 was set to /home and was designated at 20 gig, swap was designated for sda3 at 1 gb, sda4 was extended and contained sda5 which was / at 17 mb+. The attachment with mtab, fstab probably already reveal that inforation. When trying to do the installation the first time, sda2 hung for quite a while trying to create the 54th inode. After it was created, though with errors, the other partitions were created and formatted. The creations of the other partitions seemed to be a slow process also. After this failure, I decided to try the installation again, I decided to bypass the usage of sda2 as home and used the same scheme as mentioned above. (sda2 not used at all). Since the partitions were formatted already, I skipped reformatting the partitions again. When the installation ceased, I tried to recover the information from screen 2 shell. Whatever I tried to copy turned out to be 16 mb of binary data. I did not get the information that was displayed on in the terminal. Post-failure, I booted up one of my working installations and expected to be able to retrieve information in the /root directory from anaconda. There was only what is in the zip above. Checking the partitions that were created from the installer, sda2 was indeed corrupted fsck stated that the reported size was larger than the available size. the other partitions seemed to fsck alright. (didn't check swap). After mkfs -j /dev/sda2 was performed using a working installation on the same computer, the disk formatted and there was no long delays during formatting. I copied some files to the reformatted sda2 partition as well as the sda1 and sda5 partitions and they seemed to be alright. Sorry I could not capture the logs or messages from this installation attempt. I expect to encounter broken in different ways. :-)
correction, sda5 was designated as 17 gb, not mb for / Also, i noted that using the fileroller in the GUI did not show what is in the gz file in its entirety. The gz was created using mc (midnight commaander) utilities available from the F2 menu. When using mc to view the gz, it mirrors what was shown on the actual disk. (sda5)
*** Bug 151877 has been marked as a duplicate of this bug. ***
From Rawhide through FC4T1 the installer seemed to want to bail out when attempting to complete a fresh install. I had 256MB/233 MHz of memory and the installer would bail out. I replaced the module with a 512MB/333 MHZ module and I did not get the same dropout from anaconda. A symptom in the shell would be that all processes seemed to be waiting and erratically, running ls would occasionally give segfaults. The installer would allow you to select packages, then would bailout on the second or third package that was being installed. Text mode would also exhibit the same symptom of getting to the second or third package before the program bailed out. When the rawhide version bailed out, it exhibited the above listed symptoms described. The FC4T1 just seemed to exit and did not lockup the system. An install image was also created in /home of around 60 MB (rhinstall-stage2.img) The remaining filesystem and logs is probably around 9MB with all the remaining directories installed before failure. I still have the zipped filestructure handy (60MB)and the 256MB module which exhibited the error reported in bug 151877 and this earlier bug. This failure seems to only effect new installs. Upgrades seemed to be alright.
With the kernel debugging turned off, this should be better. Can you try again with test3?
I assume you mean to try this with only 256 MB installed and attempting a fresh install. I will give test3 a try to see how things proceed.
The installation did not exit while installing fresh with the FC4T3 installation discs. The install tried was an everything install and nothing bombed out.