Description of problem: Unable to configure LVM on RAID in ks.cfg file. Version-Release number of selected component (if applicable): FC5 test 1 How reproducible: Always Steps to Reproduce: ks.cfg section: part raid.11 --onpart=hda1 part raid.12 --onpart=hdc1 raid swap --fstype=swap --level=1 --device md1 raid.11 raid.12 I have tried the following raid options: #raid swap --fstype=swap --level=RAID1 --device md1 raid.11 raid.12 #raid swap --fstype=swap --level=RAID1 --device=md1 raid.11 raid.12 #raid swap --fstype=swap --level=1 --device=md1 raid.11 raid.12 #raid swap --fstype swap --level=RAID1 --device md1 raid.11 raid.12 This did work for FC4: part raid.11 --onpart=hda1 part raid.12 --onpart=hdc1 raid swap --fstype=swap --level=1 --device=md1 raid.11 raid.12 Actual results: I use 'text' mode in ks.cfg to install. Anaconda errors after mouse & screen have been probed. Error messages go scrolling off the screen too quick to catch 'em all. But do see something refering to: DoRaid. My ks.cfg file works fine without RAID. Expected results: LVM on RAID installation. Additional info: I've also refered to: http://www.redhat.com/docs/manuals/enterprise/RHEL-4-Manual/sysadmin-guide/s1-kickstart2-options.html I also tried: system-config-kickstart on FC5 test1 and got error: # I/O warning : failed to load external entity "/usr/share/comps/i386/comps.xml" Could not start because there is no /usr/share/comps/i386/comps.xml file. Please make sure the comps package is installed. I could not find comps RPM. So copied file from FC5 CD1 and then got error message: # Traceback (most recent call last): File "/usr/share/system-config-kickstart/system-config-kickstart.py", line 82, in ? kickstartGui.kickstartGui(file) File "/usr/share/system-config-kickstart/kickstartGui.py", line 165, in __init__ self.fillData() File "/usr/share/system-config-kickstart/kickstartGui.py", line 311, in fillData self.X_class.fillData() File "/usr/share/system-config-kickstart/xconfig.py", line 274, in fillData if self.ksdata.xconfig["monitor"] != "": KeyError: 'monitor'
You can find the errors anaconda spits out in /tmp/anaconda.log and sometimes also /tmp/anacdump if you get a traceback. Please attach those to this bug report and I'll be able to fix the problem.
Hi Thanks for coming back to me. I don't think the installer gets far enough to allow me to get to a shell prompt. A shell appears on [alt]+[f2], but i'm unable to type any commands :( Another reason is that, the installer 'exited abnormally' and unmounted all files systems. Thus unable to see any files in /tmp. Is there another way of getting at these files? Are you able to create RAID and LVM on RAID in kickstart? Regards W Agtail
After the exception dialog pops up and prompts you to save it to a floppy or remote machine, just don't hit any of the buttons or press enter. As long as the exception dialog is up, you should be able to switch over to tty2 and get to the log files from there. I have not tested LVM on RAID recently, though I know that RAID through kickstart should be working.
Also, could you attach your whole kickstart config file (minus any passwords, etc.)?
Created attachment 122854 [details] kickstart file - ks.cfg My ks.cfg file - refer to 'LVM on RAID' section - thanks
Hi, as requested, please find attached my ks.cfg file. I didn't get any exception dialog prompts :( Many thanks, W Agtail
The biggest problem I see is that you don't have any root partition defined. You have a couple lines in there, but they are commented out. That'd certainly cause anaconda to halt. I can't reproduce your problem on rawhide, even with LVM on RAID setups.
I can confirm that this is still a problem in FC5 test 3. I have a similar ks.cfg and hit this problem. I looked through the code and it appears that anaconda does not accept "swap" after the "raid" command. Sorry I can't get a traceback -- the installed locks up after I get the error. In brief, what I am seeing is that kickstart.py line 459 (doRaid) doesn't handle "swap" as the mountpoint. It expects the first character to be a "/" or the string to be empty. On a sidenote, I tried to work around this problem by making my swap partitions non-raid (just create swap with "part"). This caused other more confusing problems for the raid config. I'll try to recreate that problem and post a seperate bz with tracebacks.
It definitely seems that having a line in ks.cfg such as ... raid swap --fstype=swap --level=1 --device=md1 raid.11 raid.12 will not work for the reasons mentioned above. Note that this did work for FC4, and in my humble opinion it should still be valid. I attempted to move the swap to a regular "part" command in ks.cfg. For example: part swap --fstype=swap --onpart=sda part swap --fstype=swap --onpart=sdb At the same time all of my regular partitions (with ext3 as fstype) were configured as raid. Each specified the md device (--device=md#). This caused some strange behavior, namely that the installed would fail to format md1. I can recreate this. Once this happened I would go to the shell and get some info using mdadm. It appeared that md1 was not being used. For example: raid /boot --fstype=ext3 --level=1 --device=md0 raid.10 raid.11 raid / --fstype=ext3 --level=0 --device=md1 raid.12 raid.13 raid /usr --fstype=ext3 --level=0 --device=md2 raid.14 raid.15 raid /var --fstype=ext3 --level=0 --device=md3 raid.16 raid.17 raid /tmp --fstype=ext3 --level=0 --device=md4 raid.18 raid.19 raid /home --fstype=ext3 --level=0 --device=md5 raid.20 raid.21 The above raid config would not work, failing each time while trying to format md1. Note that my swap partitions were configured with "part" as stated above. My work-around was to not use md1, ie. replace md1 (above) with md2 and bump all the md# by one. I'm not sure what md1 is causing problems. It should be valid. I think the best solution is to allow swap to be a "mount point" for the raid command, as in FC4.
Committed a fix to Rawhide. Please test with the next anaconda build.
Hi I'm testing again with FC5-test3-x86_64 and still see the same problems. I'm not sure how to access Rawhide?
Is this better in FC5 final?
Closing due to no response by reporter. If this still happens in FC5 final, please reopen.
Hi, I've been away for while and just trying this with final release of FC5. I'm still seeing problems. section of ks.cfg file looks like this: bootloader --location=partition --driveorder=sda,sdb --append="rhgb quiet" reboot # # LVM on RAID # # swap mirrored part raid.11 --onpart=sda1 part raid.12 --onpart=sdb1 raid swap --fstype=swap --level=1 --device=md1 raid.11 raid.12 # /boot mirrored part raid.21 --onpart=sda2 part raid.22 --onpart=sdb2 raid /boot --fstype=ext3 --level=1 --device=md2 raid.21 raid.22 # /root mirrored part raid.31 --onpart=sda3 part raid.32 --onpart=sdb3 raid / --fstype=ext3 --level=1 --device=md3 raid.31 raid.32 # LVM part raid.51 --ondisk=sda --size=1 --grow part raid.52 --ondisk=sdb --size=1 --grow raid pv.5 --fstype "physical volume (LVM)" --level=1 --device=md5 raid.51 raid.52 volgroup sys --pesize=32768 pv.5 logvol /usr --vgname=sys --size=10240 --name=usr --fstype=ext3 logvol /usr/local --vgname=sys --size=1024 --name=usrlocal --fstype=reiserfs logvol /opt --vgname=sys --size=1024 --name=opt --fstype=reiserfs logvol /tmp --vgname=sys --size=1024 --name=tmp --fstype=reiserfs logvol /var --vgname=sys --size=5120 --name=var --fstype=ext3 logvol /var/www --vgname=sys --size=2048 --name=varwww --fstype=ext3 logvol /home --vgname=sys --size=8192 --name=home --fstype=ext3 logvol /audio --vgname=sys --size=14336 --name=audio --fstype=reiserfs logvol /downloads --vgname=sys --size=4096 --name=downloads --fstype=ext3 logvol /scratch --vgname=sys --size=1 --name=scratch --fstype=reiserfs --grow %packages # Refer to: egrep "<id>|<name>|<description>|<packagereq|<subcategory>" /usr/share/comps/x86_64/comps.xml # # Minimum Install # @ admin-tools @ base @ base-x @ core @ development-tools @ dialup @ editors @ gnome-desktop @ graphical-internet @ sound-and-video kernel lvm2 reiserfs-utils system-config-kickstart # # Install almost everything. # #@ Everything %pre # Trash partition table on sda (primary) and sdb (mirror). # Add partitions ready for swap(2G), /boot(128M) and /(1G). parted -s /dev/sda mklabel msdos parted -s /dev/sda mkpart primary 0 2048 parted -s /dev/sda mkpart primary 2048 2176 parted -s /dev/sda mkpart primary 2176 3200 parted -s /dev/sdb mklabel msdos parted -s /dev/sdb mkpart primary 0 2048 parted -s /dev/sdb mkpart primary 2048 2176 parted -s /dev/sdb mkpart primary 2176 3200 %post echo "# Version 1.1
Created attachment 131083 [details] anacdump.txt
Created attachment 131084 [details] anaconda.log
Just to confirm, this ks file works ok with FC4.
FYI: I just tried today's rawhide version from: http://download.fedora.redhat.com/pub/fedora/linux/core/development/x86_64/ and get the same results :( thanks, Wagtail
This should be fixed in Rawhide after FC6 test1.
I've tried several revisions of rawhide. For example today's rawhide breaks at kickstart. The process seems to 'blue screen'. doesn't get as far as a shell session. i'm installing using PXE and another another FC computer for DHCP etc.
If you are seeing problems with anaconda besides this bug, please open different bugs against anaconda. Otherwise, please only reopen this bug if you are continuing to see this same problem.