Bug 189971
|
Description
Bob Gustafson
2006-04-26 11:35:12 UTC
Can you grab /tmp/anaconda.log and /tmp/syslog from within the installer? How do I do that? I can do it with 'linux rescue' - send through network, but I don't know how to do it in the installer. There I just click on the buttons - not much in the way of command line interaction. Or is there a way? If you switch to tty2, you'll have a shell and can bring up the network or copy to removable storage. Hmm, might I use the shell to 'mdadm --create ...' etc? I looked in anaconda.log and syslog. Seems to be something left over from install run. Is this possible. Should I wipe what is there and start anew from the installer and tty2 ? OK, did the whole job, but am having same bum results. Starting from the beginning 0) selecting 'linux text' mode (may not make a difference) 1) boot with cdrom #1 up to the step where it welcomes the user, just after the English and keyboard (us) question (early in the process anyway) 2) Go into alternate screens ALT F1 - gets back to the installer screen ALT F2 - gets a shell (tty2). Don't say exit, because you don't come back ALT F3 - a tail view of /tmp/anaconda.log ALT F4 - a tail view of /tmp/syslog 3) In the tty2 screen, do the following /usr/sbin/mdadm --create /dev/md0 -l 1 -n 2 /dev/sdb1 /dev/sdc1 /usr/sbin/mdadm --create /dev/md1 -l 1 -n 2 /dev/sdb2 /dev/sdc2 /usr/sbin/mdadm --create /dev/md2 -l 1 -n 2 /dev/sdb3 /dev/sdc3 4) Check for completion of md[n] assemble (takes about 20 minute) by executing the following command until the display shows that all 3 arrays are active cat /proc/mdstat 4) Check/create the LVM phys vol, group, and logical volume execute: /usr/sbin/lvm File descriptor 101 left open --- (garbage output?) File descriptor 107 left open --- (garbage output?) lvm> pvdisplay ... good looking output, includes VG Name rootvg If volume shows pv already created, then go on, otherwise create p volume --- conditional lvm> pvcreate /dev/md2 lvm> vgdisplay --- good looking output, includes VG Status resizable If volume shows rootvg already create, then go on, otherwise create Now, check for logical volume lvm> lvdisplay --- goog looking output, includes LV Name /dev/rootvg/root Note, if this output includes LV Status NOT available, then you need to do lvm> vgchange -a y and check again. lvm> lvdisplay --- good looking output, including LV Status available 5) Now, switch back to the Installer screen (ALT F1) and continue answering the questions (English, keyboard )? 6) You will see that when you come to the 'Partitioning Type' screen, only the sdb disk is shown - you cannot go on from here because you only have one disk. Need two to RAID. 7) Jumping back to the ALT F2 screen and executing lvm> lvdisplay You (I anyway) see that the /dev/rootvg/root has changed status to: LV Status NOT available (Curious..) anaconda.log and syslog are given as attachments to follow (These were copied off using rsync, once the eth0 interface was configured by hand with /usr/sbin/ifconfig) Created attachment 128281 [details]
anaconda.log - after all of the actions described above
Created attachment 128283 [details]
syslog -- after all of the actions described above
Is one of these a firewire or USB disk? We don't support installing to them by default... There may have been a firewire disk attached to the box, but sometime in that process my keyboard fell off the tower and hit the firewire box. It started to make a funny sound, so I turned it off. I don't know whether that was before or after the attached logs. In either case, firebox was not used. There was no storage attached to a USB port. At one time in my earlier experiments, I attached my camera via USB. I should have been able to see the memory (0.5Gb), but did not. Camera was not attached for above logs. The two disks sdb and sdc are 73Gb Seagate SCSI disks. Identical numbers of blocks in each. The sdc disk originally had rh9 installed. When I wanted to go back to rh9 from fc[2345] I would just unplug both disks and swap them in their SCA slots. I could then mount the fc disk from rh9. The idea of getting the second disk was to go to RAID1. Because I needed two (blank) disks to go to RAID at fc install time, this only happened when I got another (firebox) 200G disk to copy stuff to. Right now, this whole .. machine is dead in the water. ----- I do have another machine, much more modest, with two 36G disks running RAID1 FC5. I had a third 4G scsi disk on that machine which was (is?) the grub host. So I know what I am trying to do is possible.. Well, that's pretty interesting. If I boot up the machine using Fedora Core FOUR disc 1, going through the English, English us, to the Disk Setup window: Both SCSI disks show (sda, sdb), the LVM Volume Group rootvg->root is shown, all 3 RAID Devices are shown /dev/md0, /dev/md1, /dev/md2 and the rootvg is shown on the Mount Point/RAID/Volume column of /dev/md2, just as it should be.. A difference between this and FC5 is that sda on FC5 seems to be the cdrom drive (maybe the loop device - neet to go back and look). The first (and only) SCSI disk in the FC5 install is sdb - it does not see the other disk, and therefore cannot start up the RAID devices. I think I will go forward and install FC4 on this system and then go back and try to install FC5. Perhaps something will have changed (??). So far, so good. Installed FC4 and am running from it now. Note this is FC FOUR, not the latest and greatest FC five. Attached are the dmesg, anaconda-ks.cfg, install.log, and install.log.syslog for this installation on this hardware of FC4 Created attachment 128414 [details]
dmesg taken just after initial login by user1 on Fedora FOUR
This should show the same hardware as shown in the FC5 attempts. Note that the
firewire disk is still turned OFF.
Created attachment 128415 [details]
install.log from install of Fedora FOUR on this hardware
Created attachment 128416 [details]
install.log.syslog for Fedora FOUR install on same hardware
The attach process for anaconda-ks.cfg said file was empty. It actually has 1835 bytes. This was after copying to /home/user1 and chown user1:user1 Directory is shown below: [root@hoho2 ~]# ls -l total 124 -rw-r--r-- 1 user1 user1 1835 Apr 30 11:56 anaconda-ks.cfg -rw-r--r-- 1 user1 user1 25472 Apr 30 11:57 dmesg -rw-r--r-- 1 user1 user1 61933 Apr 30 11:34 install.log -rw-r--r-- 1 user1 user1 5541 Apr 30 11:34 install.log.syslog [root@hoho2 ~]# One more of the install puzzles.. I do like the ALT F2, F3, F4 of the new FC5 installer. -------- Now for retry of FC5 install.. However, the ALT F2, F3, F4 seems to be only available if one does a 'linux text' install. I have successfully installed FC5 on RAID1, LVM as I wanted, but I don't have any corresponding anaconda.log and syslog. So, I will do the install over again, but with 'linux text' to get those files for you. It is a pretty quick install because I deselected the checkmark on all of the system types (office, developer, everything, whatever..) -- Also the sound test asks for the installer to click on the 'Play Test Sound' button. There is no button of that name, only an arrow (play?) button. Also, before I wipe the files from this FC5 install, the bottom of the install.log.syslog file says something about Network is unreachable. This must be at an earlier stage of the install, because the network is certainly reachable, see screen grab below: --- This is the bottom of cat /root/install.log.syslog from FC5 install <86>Apr 30 12:54:39 groupadd[3218]: new group: name=beagleindex, GID=58 <86>Apr 30 12:54:39 useradd[3219]: new user: name=beagleindex, UID=58, GID=58, home=/var/cache/beagle, shell=/bin/false <86>Apr 30 13:03:26 useradd[4226]: new group: name=gdm, GID=42 <86>Apr 30 13:03:26 useradd[4226]: new user: name=gdm, UID=42, GID=42, home=/var/gdm, shell=/sbin/nologin <86>Apr 30 13:03:26 usermod[4227]: change user `gdm' shell from `/sbin/nologin' to `/sbin/nologin' <11>Apr 30 13:04:42 python: failed to set default route: Network is unreachable [root@hoho2 ~]# ping 192.168.49.21 PING 192.168.49.21 (192.168.49.21) 56(84) bytes of data. 64 bytes from 192.168.49.21: icmp_seq=1 ttl=64 time=5.14 ms 64 bytes from 192.168.49.21: icmp_seq=2 ttl=64 time=0.304 ms 64 bytes from 192.168.49.21: icmp_seq=3 ttl=64 time=0.345 ms 64 bytes from 192.168.49.21: icmp_seq=4 ttl=64 time=0.389 ms --- 192.168.49.21 ping statistics --- 4 packets transmitted, 4 received, 0% packet loss, time 3012ms rtt min/avg/max/mdev = 0.304/1.546/5.149/2.080 ms [root@hoho2 ~]# ping www.google.com PING www.l.google.com (72.14.203.104) 56(84) bytes of data. 64 bytes from 72.14.203.104: icmp_seq=1 ttl=245 time=13.0 ms 64 bytes from 72.14.203.104: icmp_seq=2 ttl=245 time=12.0 ms 64 bytes from 72.14.203.104: icmp_seq=3 ttl=245 time=11.1 ms --- www.l.google.com ping statistics --- 3 packets transmitted, 3 received, 0% packet loss, time 2001ms rtt min/avg/max/mdev = 11.137/12.076/13.006/0.763 ms [root@hoho2 ~]# Well, that's pretty interesting. If I boot up the machine using Fedora Core FOUR disc 1, going through the English, English us, to the Disk Setup window: Both SCSI disks show (sda, sdb), the LVM Volume Group rootvg->root is shown, all 3 RAID Devices are shown /dev/md0, /dev/md1, /dev/md2 and the rootvg is shown on the Mount Point/RAID/Volume column of /dev/md2, just as it should be.. A difference between this and FC5 is that sda on FC5 seems to be the cdrom drive (maybe the loop device - neet to go back and look). The first (and only) SCSI disk in the FC5 install is sdb - it does not see the other disk, and therefore cannot start up the RAID devices. I think I will go forward and install FC4 on this system and then go back and try to install FC5. Perhaps something will have changed (??). I am using two machines - the target machine, where I inserted a comment just recently, and this machine where I submitted a comment awhile ago. When I went back to this machine and did a browser reload, I received the dialog box saying there was a collision and if I wanted to overwrite, or loose something. I chose the overwrite, and I see that the text immediately above, which was submitted at 08:24 today, is submitted again. What was lost? Who knows? The text above at 14:57 seems to be the last text submitted from my other (target) machine. Moving along now.. I went back to the Fedora Core FIVE disc 1 and did a 'linux text' install. I actually did not do the whole install, because all I wanted was the anaconda.log and syslog for the FC5 install which includes correct information (both SCSI disks detected at this early stage. Attachments below are anaconda.log and syslog grabbed after the language and keyboard questions and before anything is written on the disks. Created attachment 128421 [details]
anaconda.log for correct install run of FC5
Created attachment 128422 [details]
syslog from correct install run of FC5
I'm done. I have a good FC5 install and am now downloading updates and then will download additional software and then will configure some of the things that don't get configured (vncserver, ..) I will also write up a suggestion for a better way to install Fedora (minium software on install, then add additional software from CD's and/or web after the basic Fedora is installed). This scheme would make it easier for you to debug, because there is only one minimum system being installed.. See you.. Going over some of my own questions: 1) Booting up the installer in 'linux text' mode with the Firewire box turned off, gives a partition display which DOES NOT include a non scsi sdba. The two SCSI disks are now labeled sda and sdb - there is no sdc as in the previous (early in this note) display sh-3.1# cat /proc/partitions major minor #blocks name 7 0 70420 loop0 8 0 71687372 sda 8 1 104391 sda1 8 2 3911827 sda2 8 3 67665780 sda3 8 16 71687372 sdb 8 17 104391 sdb1 8 18 3911827 sdb2 8 19 67665780 sdb3 Now, to turn on firewire disk and see if this is the thing that screws up the subsequent partitioning and RAID LVM in the install process.. OK, that was the problem !! If the firewire disk is turned on and plugged into the machine, then it is listed in cat /proc/partitions as sda. The two SCSI disks are sdb and sdc. AND, when I go further into the Fedora Install, the sdc is not seen by the installer. Booting the FC5 disc 1 and choosing the 'linux text' mode, all three disks are shown (sda (firew), sdb (SCSI 1), and sdc (SCSI 2) in the ALT F2, cat /proc/partitions list. This was the original complaint. So, the conclusion is that having a firewire disk connected to your machne while you try to install Fedora Core 5 will screw up the installation process. It can be fixed by just unplugging the firewire disk. Done. Closing as NOTABUG on the basis of comment 27. Firewire in Linux is "interesting" anyway, so I can see there being problems here with disks moving around. Thanks for doing the research on this to figure out what your problem was. The research was mostly frustration with a lucky accident https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=189971#c9 I would say that it is still a bug. The underlying problem (limited media visibility by anaconda) is still there. At the minium, to fix bug, please beef up install documentation to say that all strange media must be disconnected and all RAID arrays must not be in a degraded state prior to attempting install of Fedora. Fixed in CVS for F7 IG. |