Hide Forgot
Description of problem: We just started seeing this in rhel8 while setting up PVs on top of luks volumes. 10 disk(s) to be used: hayes-01=/dev/sdd /dev/sdc /dev/sdb /dev/sdj /dev/sde /dev/sdh /dev/sdi /dev/sdg /dev/sdk /dev/sdf on hayes-01... dicing /dev/sdd into 1... dicing /dev/sdc into 1... dicing /dev/sdb into 1... dicing /dev/sdj into 1... dicing /dev/sde into 1... dicing /dev/sdh into 1... dicing /dev/sdi into 1... dicing /dev/sdg into 1... dicing /dev/sdk into 1... dicing /dev/sdf into 1... re-reading disks on hayes-01... Zeroing out the new partitions.../dev/sdd1.../dev/sdc1.../dev/sdb1.../dev/sdj1.../dev/sde1.../dev/sdh1.../dev/sdi1.../dev/sdg1.../dev/sdk1.../dev/sdf1... Creating crypt volumes on the following storage: /dev/sdd1 /dev/sdc1 /dev/sdb1 /dev/sdj1 /dev/sde1 /dev/sdh1 /dev/sdi1 /dev/sdg1 /dev/sdk1 /dev/sdf1 cryptsetup luksFormat /dev/sdd1 cryptsetup luksOpen /dev/sdd1 cPV1 cryptsetup luksFormat /dev/sdc1 cryptsetup luksOpen /dev/sdc1 cPV2 cryptsetup luksFormat /dev/sdb1 cryptsetup luksOpen /dev/sdb1 cPV3 cryptsetup luksFormat /dev/sdj1 cryptsetup luksOpen /dev/sdj1 cPV4 cryptsetup luksFormat /dev/sde1 cryptsetup luksOpen /dev/sde1 cPV5 cryptsetup luksFormat /dev/sdh1 cryptsetup luksOpen /dev/sdh1 cPV6 cryptsetup luksFormat /dev/sdi1 cryptsetup luksOpen /dev/sdi1 cPV7 cryptsetup luksFormat /dev/sdg1 cryptsetup luksOpen /dev/sdg1 cPV8 cryptsetup luksFormat /dev/sdk1 cryptsetup luksOpen /dev/sdk1 cPV9 cryptsetup luksFormat /dev/sdf1 cryptsetup luksOpen /dev/sdf1 cPV10 creating lvm devices... hayes-01: pvcreate /dev/mapper/cPV10 /dev/mapper/cPV9 /dev/mapper/cPV8 /dev/mapper/cPV7 WARNING: atari signature detected on /dev/mapper/cPV8 at offset 466. Wipe it? [y/n]: [n] Aborted wiping of atari. 1 existing signature left on the device. WARNING: atari signature detected on /dev/mapper/cPV7 at offset 466. Wipe it? [y/n]: [n] Aborted wiping of atari. 1 existing signature left on the device. couldn't pvcreate /dev/mapper/cPV10 /dev/mapper/cPV9 /dev/mapper/cPV8 /dev/mapper/cPV7 on hayes-01 Version-Release number of selected component (if applicable): cryptsetup-2.0.6-1.el8 BUILT: Mon Dec 3 07:46:20 CST 2018 cryptsetup-libs-2.0.6-1.el8 BUILT: Mon Dec 3 07:46:20 CST 2018 cryptsetup-reencrypt-2.0.6-1.el8 BUILT: Mon Dec 3 07:46:20 CST 2018 kernel-4.18.0-49.el8 BUILT: Tue Dec 4 06:01:31 CST 2018 lvm2-2.03.01-1.el8 BUILT: Thu Nov 1 04:29:08 CDT 2018 lvm2-libs-2.03.01-1.el8 BUILT: Thu Nov 1 04:29:08 CDT 2018 lvm2-dbusd-2.03.01-1.el8 BUILT: Thu Nov 1 04:31:09 CDT 2018 lvm2-lockd-2.03.01-1.el8 BUILT: Thu Nov 1 04:29:08 CDT 2018 device-mapper-1.02.153-1.el8 BUILT: Thu Nov 1 04:29:08 CDT 2018 device-mapper-libs-1.02.153-1.el8 BUILT: Thu Nov 1 04:29:08 CDT 2018 device-mapper-event-1.02.153-1.el8 BUILT: Thu Nov 1 04:29:08 CDT 2018 device-mapper-event-libs-1.02.153-1.el8 BUILT: Thu Nov 1 04:29:08 CDT 2018 device-mapper-persistent-data-0.7.6-1.el8 BUILT: Sun Aug 12 04:21:55 CDT 2018 How reproducible: Often
This may happen due to atari signature being too short and therefore often detected by mistake even with pseudo-random data on new unlocked LUKS device. Karel, do you have any idea if there's a way to improve blkid scan so that it doesn't find the signature so often? Otherwise I can only advise to wipe all /dev/mapper/cPV* devices first or just run pvcreate -y.
I think call "wipefs -a" is a good idea to avoid wrong interpretation of the random data ;-) Unfortunately, I do not see a way how to make libblkid atari probing function more robust...