Bug 1207317

Created attachment 1008607 [details]
File: anaconda-tb

Created attachment 1008608 [details]
File: anaconda.log

Created attachment 1008609 [details]
File: dnf.log

Created attachment 1008610 [details]
File: environ

Created attachment 1008611 [details]
File: lsblk_output

Created attachment 1008612 [details]
File: nmcli_dev_list

Created attachment 1008613 [details]
File: os_info

Created attachment 1008614 [details]
File: program.log

Created attachment 1008615 [details]
File: storage.log

Created attachment 1008616 [details]
File: syslog

Created attachment 1008617 [details]
File: ifcfg.log

Created attachment 1008618 [details]
File: packaging.log

Nominating as a Beta blocker: https://fedoraproject.org/wiki/Fedora_22_Beta_Release_Criteria#Hardware_and_firmware_RAID , "The installer must be able to detect and install to hardware or firmware RAID storage devices."

scratch build with a potential fix:
http://koji.fedoraproject.org/koji/taskinfo?taskID=9368091

Created attachment 1008635 [details]
logs from install attempt with the scratch build

With the scratch build it no longer crashes, but the set does not appear as an available installation target (so if your only target is a RAID set, you can't install).

Attaching the logs (all smooshed into one file by fpaste).

Discussed at 2015-03-30 blocker review meeting: http://meetbot.fedoraproject.org/fedora-blocker-review/2015-03-30/f22-blocker-review.2015-03-30-16.04.log.txt . Accepted as a blocker per criterion cited in #c13.

So the key is this (from storage.log):
17:34:05,125 DEBUG blivet:           DeviceTree.handleUdevDeviceFormat: name: sda ;
17:34:05,136 INFO blivet: type detected on 'sda' is 'isw_raid_member'
17:34:05,138 DEBUG blivet:             MDRaidMember.__init__: uuid: None ; exists: True ; label: None ; device: /dev/sda ; serial: ST3500418AS_9VM1BT6B ; mdUuid: None ; biosraid: True ;
17:34:05,138 DEBUG blivet: getFormat('isw_raid_member') returning MDRaidMember instance with object id 10
17:34:05,140 DEBUG blivet:             DiskDevice._setFormat: sda ; current: None ; type: mdmember ;
17:34:05,140 INFO blivet: got format: existing mdmember
17:34:05,143 DEBUG blivet:            DeviceTree.handleUdevMDMemberFormat: type: mdmember ; name: sda ;
17:34:05,162 DEBUG blivet:             DeviceTree.getDeviceByUuid: hidden: False ; uuid: None ; incomplete: True ;
17:34:05,164 DEBUG blivet:             DeviceTree.getDeviceByUuid returned None
17:34:05,164 ERR blivet: failed to determine name for the md array e0fb362e-9865-d1c0-cdb0-f5e327c9a085

and the problem behind that is that this:
17:34:05,153 INFO program: Running [5] mdadm --examine --brief /dev/sda ...
17:34:05,160 INFO program: stdout[5]: ARRAY metadata=imsm UUID=e0fb362e:9865d1c0:cdb0f5e3:27c9a085
ARRAY /dev/md/Volume0 container=e0fb362e:9865d1c0:cdb0f5e3:27c9a085 member=0 UUID=1d1335e9:79d4be6a:e16b7fa7:fe8c41f6

in case of a SW MD RAID array starts with 'ARRAY /dev/md/something' which libblockdev uses to determine the name of the device (and so used to do blivet earlier).

I cannot imagine how this worked before, but if this works with F22 Alpha (no libblockdev), we need to investigate it more. Otherwise libblockdev can just search for the 'ARRAY /dev/md/something' pattern in the whole output to get the name.

The last time I know for sure it worked was 21 Final. I'll check 22 Alpha later and attach logs if it works.

So with some changes in libblockdev (I hesitate to call them fixes now) [1] we get to bug #1208536.

[1] http://vpodzime.fedorapeople.org/1207317_updates.img

Pull request with the libblockdev changes:
https://github.com/rhinstaller/libblockdev/pull/7

re-assigning to libblockdev for clarity.

Booting TC6 with the updates.img from #c19 I do indeed get to #1208536 , but if I build current libblockdev package plus https://github.com/rhinstaller/libblockdev/pull/7.patch and include that libblockdev build in a boot.iso , I still get 'malformed or invalid UUID'. I think some fix from the updates.img must be missing from the PR.

(In reply to awilliam from comment #22)
> Booting TC6 with the updates.img from #c19 I do indeed get to #1208536 , but
> if I build current libblockdev package plus
> https://github.com/rhinstaller/libblockdev/pull/7.patch and include that
> libblockdev build in a boot.iso , I still get 'malformed or invalid UUID'. I
> think some fix from the updates.img must be missing from the PR.
As explained in the PR, the change related to UUIDs was already pushed to master and thus is not part of the PR (I considered the change safe and clearly right).

oh, sorry, missed that. it made testing a bit tricky as there isn't an image with those changes :) i figured out a way, though.

libblockdev-0.9-1.fc22, python-blivet-1.0.7-1.fc22, anaconda-22.20.9-1.fc22 has been submitted as an update for Fedora 22.
https://admin.fedoraproject.org/updates/libblockdev-0.9-1.fc22,python-blivet-1.0.7-1.fc22,anaconda-22.20.9-1.fc22

Fix verified with a test image built with the bits from #c25.

*** Bug 1209263 has been marked as a duplicate of this bug. ***

Update has gone stable, closing (I'll verify FW RAID with Beta RC2 soon).