The latest version of lvm2 appears to rely on udev. Last messages in ddebug.log are * looking for vgs in sda1 * running: lvm2 vgscan Reading all physical volumes. This may take a while... WARNING: Device /dev/ram0 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/sda not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/vg-mga/lv_mga1 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/ram1 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/sda1 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/vg-mga/lv_mga2 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/ram2 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/sda2 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/ram3 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/sda3 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/ram4 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/sda4 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/ram5 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/sda5 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/ram6 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/sda6 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/ram7 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/sda7 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/ram8 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/sda8 not initialized in udev database even after waiting 10000000 microseconds. WARNING: Device /dev/ram9 not initialized in udev database even after waiting 10000000 microseconds. When I run 'lvm2 vgscan' manually in a chroot, it goes through all the devices, then loops back and starts trying again. @Thierry, advice please. This is stopping me building Live ISOs on the Mageia infra. Do we really need to run the setupSCSI step for a local install?
Created attachment 10707 [details] do not call 'setupSCSI' for local install Can you try this patch?
Created attachment 10708 [details] alternative patch
Created attachment 10709 [details] Extended alternative patch The alternative version cures the initial problem, but you then get a "Oops, no root partition" from check_hds_boot_and_root(). This revised patch fixes that, but maybe there's a better way?
Attachment 10708 is obsolete: 0 => 1
Use the first patch then
(In reply to Thierry Vignaud from comment #4) > Use the first patch then No, that doesn't help. It hangs in the call to install::any::getHds($o). And not calling that leads to the "Oops, no root partition".
Created attachment 10713 [details] Extended alternative patch v2 This simplified patch works for the Live ISO build.
Attachment 10709 is obsolete: 0 => 1
Unless you object, I'll push my fix.
Your patch alter the installer semantics. eg: $o->{fstab} is no more used we don't look anymore for mounted device in $o->{fstab} $o->{fstab} is overwritten & reduced to only the partition Note that we care much in drakx-in-chroot case but it would be better to have a nice explaining commit log
If we skip the setupSCSI step, $o->{fstab} is empty, so we aren't overwriting anything, and there are no other partitions to mount. As far as I can tell, we just need that one fake fstab entry. Proposed commit message: ==== Skip setupSCSI step when run with --local_install (mha#24278) In a local install, we don't have udev running, so the setupSCSI step will hang if it tries to probe for logical volumes (lvm2 uses udev). A local install is used to test the installer(drakx_in_chroot) and to build the Live ISOs (draklive2), and in both cases we don't really want the install to be affected by the hardware of the host system. Skipping the setupSCSI step means $o->{fstab} contains no entries, so we add a fake entry for our chroot, to allow us to pass the subsequent check that we have a root partition. ==== BTW, this should also fix bug 24201.
Good enough for me. Go ahead with godspeed
Thanks. I've also added a comment in the code, as a reminder.
Confirmed fixed when running Live ISO build on rabbit.
Resolution: (none) => FIXEDStatus: NEW => RESOLVED