Red Hat Bugzilla – Bug 1167014
Manual partitioning using single partition: unable to escape from Partitioning page except with workaround
Last modified: 2014-12-07 18:46:09 EST
Description of problem:
Just tried installing F21_TC3 - everything on single partition, without creating swap.
I have seen a warning stripe about no swap but pressing Done several times did not allow me to complete the screen.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. try to install without swap partition
2. press Done
3. warning is presented instructing user to press Done again to force install
Pressing Done does not proceed with installation
Pressing Done should proceed with installation
IMO failing Final Blocker criterion:
The installer must be able to create and install to any workable partition layout using any file system and/or container format combination offered in a default installer configuration.
Everything works correctly for me - if I press Done for the second time, I get a confirmation dialog and I'm returned to the main hub. Please provide exact reproduction steps (or a video) and logs, thanks.
(Tested with TC3 as well, Server DVD)
Installed using netinst image and custom partitioning. I tried twice with TC3 - both real HW and VM. Using partitions only. Attaching logs from the VM install.
Created attachment 960786 [details]
Created attachment 960787 [details]
Created attachment 960788 [details]
Created attachment 960789 [details]
Created attachment 960790 [details]
the rest of /var/log/anaconda/ files
Can you also append exact steps to reproduce?
What I tried:
1. enter custom part with an empty disk
2. click "create layout automatically"
3. delete swap
4. confirm (Done twice)
1. enter custom part with an empty disk
2. create a root partition (nothing else)
3. confirm (Done twice)
All done using LVM (the default).
- single empty disk (thin-provisioned - should be all zeros, no MBR)
- using virt-manager to create new VM with current TC3 netinstall iso
In installation destination - Device Selection
1. pick "I will configure partitioning"
In manual patitioning:
3. +: Mountpoint "/", size "8G". Add mount point. (The only mountpoint.)
4. Device type to standard patition
6. goto 5.
After adding /boot it works as expected and the configuration is accepted after pressing Done second time.
When I remove /boot now it still works.
Same happens when I add /home.
So looks like it does not work only when using single partition.
That may be non-standard partitioning today - I used this to check installation on a NB which did not have enough space in VG while enough to create a partition.
Alternative path: In manual patitioning:
3. Change partitioning scheme to "Standard Partition" instead of "LVM".
4. +: Mountpoint "/", size "8G". Add mount point. (The only mountpoint.)
6. goto 5.
OTOH if instead of 4 I do:
4.1 +: Mountpoint "/", size "8G". Add mount point.
4.2 +: Mountpoint "/home", size "1G". Add mount point.
4.3 -: /home
6. Warning. Press Done. and it is Done.
Creating and removing a partition changes some internal state.
I tried both approaches in comment 11 and comment 12, all of them work for me as expected.
It works for LVM. Does not for partitions.
Discussed at 2014-11-24 blocker review meeting: http://meetbot.fedoraproject.org/fedora-blocker-review/2014-11-24/f21-blocker-review.2014-11-24-17.01.log.txt . It seems that no-one but Marian can reproduce this even with the steps provided (bcl says he can't either), and even in the event that we do find a reproducer approximately matching the description, we felt it's enough of an unusual case that it would be OK to just document the 'create a partition then remove it' workaround.
If a clear reproducer can be found and confirmed it would be OK to propose this for freeze exception consideration, but we didn't want to vote on FE while it wasn't clear if we could even reliably reproduce a problem here.
This can be reproduced by adding a label to the mountpoint and not hitting 'Update Settings'. Clicking Done shows the warning, but doesn't clear the need for hitting 'Update Settings'.
(In reply to email@example.com from comment #16)
> This can be reproduced by adding a label to the mountpoint and not hitting
> 'Update Settings'. Clicking Done shows the warning, but doesn't clear the
> need for hitting 'Update Settings'.
Reproduced with TC3.
1. enter custom part
2. delete all existing partitions
3. create a single / partition covering the whole disk
4. type something into Label
5. hit Done repeatedly, nothing happens, can't escape
6. erase everything from Label
7. hit Done repeatedly, still can't escape
If you change something, Update Settings is required in order to escape the dialog. It would be nice if anaconda told you something like "you still have pending changes, please confirm them first" or some other approach.
Proposing as a freeze exception.
(In reply to Marian Csontos from comment #11)
> - single empty disk (thin-provisioned - should be all zeros, no MBR)
At the end the error gets triggered _only_ when using thin-volume as backing device. I verified the device IS all zeros. I will try to find out more.
(In reply to Kamil Páral from comment #17)
> If you change something, Update Settings is required in order to escape the
> dialog. It would be nice if anaconda told you something like "you still have
> pending changes, please confirm them first" or some other approach.
The solution is even simpler than that. It needs to reset the state of the button. When you click Done the current settings get saved as if you pushed update settings.
> Proposing as a freeze exception.
+1 -- I have a fix on master.
(In reply to Marian Csontos from comment #18)
> At the end the error gets triggered _only_ when using thin-volume as backing
> device. I verified the device IS all zeros. I will try to find out more.
This doesn't have anything to do with thin, it is because something in the right side (like label) has been changed and done is hit without hitting update setttings.
Discussed at today's blocker review meeting . Accepted as a Freeze Exception. A fix for this would be considered during freeze.
(In reply to firstname.lastname@example.org from comment #19)
> (In reply to Marian Csontos from comment #18)
> > At the end the error gets triggered _only_ when using thin-volume as backing
> > device. I verified the device IS all zeros. I will try to find out more.
> This doesn't have anything to do with thin, it is because something in the
> right side (like label) has been changed and done is hit without hitting
> update setttings.
I am the reporter and I am pretty positive the original issue I reported has something to do with thin as I were able to reproduce it 100% while Kamil were not until the label trick was invented but I am prone to believe these are two unrelated issues.
I am letting this report be hijacked for the "label" issue and I will open another bug when I have data on thin-related part and/or if still reproducible with the next TC or GA. This is a specific configuration and it will not be worth of blocker or exception.
Not clear on whether this landed in 21 or not, but it certainly landed in master, so let's close this by setting it to Rawhide.