Description of problem: When doing a text based install and creating a new raid array, the new raid array gets set to using /dev/md0 when it is actually created, even though the setup screen indicates it will be created using the correct md device. Note that this does not happen when doing graphical installs. A possible hint to the problem is that the kickstart log file will have the device name is mdNone for a text based install, while for a graphical install you will get a device name like md3. Version-Release number of selected component (if applicable): 11.1.1.3 How reproducible: It seems to happen reliably. Steps to Reproduce: 1. Have a system with at least one preinstalled (as /dev/md0) software raid partion. And another FC OS somewhere other than /dev/md0 (otherwise it will be hard to look at things later). 2. Do a fresh install and use a custom partition layout. 3. Build a new software raid device (in my tests I was deleting an existing one and then recreating it) 4. I used the new raid device as the only device in the install (other than swap) so that its mount point was / and it was set to be reformatted. 5. At this point things look good and I continue with the install. 6. Boot the system on the prexisting FC OS and examine the raid devices. Or alternatively start another install using a custom layout and note that there are two /dev/md0 arrays. Actual results: The new software raid array gets created using /dev/md0 instead of the one indicated in the setup screen. Expected results: That the raid array would be created using the correct md device. Additional info:
Fix important typo in subject.
This is because we're never setting the raidminor number in the text interface, while in the graphical interface we require you to select the minor number via a widget. Committed a fix to set a minor number automatically.
If you want I can test this. I didn't see an updated rpm in devlopment, fc6 test or fc6 updates, so I haven't tested it yet. This isn't something I really need, so if you don't need further testing, that's OK too.
We haven't built a new anaconda package yet, but we should do that today. There won't be a new anaconda in fc6-updates since anaconda is special.
Will it show up in rawhide? I have been making my own respins with a custom anaconda (which is how I happened to be looking for issues with text installs and software raid) and can use a src rpm to make a new install dvd. Again, I really don't need software raid to work with test installs for what I am doing, but I am willing to test out the fix for Fedora as probably not too many people try using software raid with text installs.
Yes, once we rebuild the anaconda package, the next day's rawhide tree will have the fix.
I don't think the fix for this was correct. I built a respin using anaconda-11.2.0.1-1 and libdchp* from rawhide and otherwise up to date fc6 rpms. I built anaconda from the src rpm under fc6 with the above dchp libs. When I tested the text based install, the new partition got a minor number, but it got the wrong one. I had existing raid partitions of md0, md1, md2, md3, md4 and md5. I deleted md3 and then assembled a new raid partition out of the freed up partitions. The new raid partition got called md5, which was still in use, rather than using md3, which the graphical install did. It's possible that not having all of rawhide caused some problems (as I will report on separately), but it looks like this problem was local to ananconda.
*** Bug 232237 has been marked as a duplicate of this bug. ***
I have a fix for this that will be in the next build of anaconda. Thanks for your patience.
I retested this with the rawhide boot.iso from Friday morning March 30. Now when I delete and recreate a raid partition during a text install the new raid device gets a name that matches the highest numbered md device instead of md0. I had existing arrays md0, md1, md2, md3, md4, md5 and md6 and deleted md5. Then the new array got named md6 (as shown in the menu, I aborted the install at that point) instead of either md5 or md7. When I tried this with a graphical install, the new array was named md5.
This will be in the next build of anaconda. Thanks.
I retested this and it does work for me now.