FYI - SoftRAID weirdness after kernel 2.6.26-2-amd64 upgrade

Post #313,145 by scoenye 8/26/09 9:26:13 PM Reply	FYI - SoftRAID weirdness after kernel 2.6.26-2-amd64 upgrade I have a box running Debian on a 2 disk software RAID 1 + LVM2 setup. A couple of days ago a kernel security update was installed, bringing it to 2.6.26-17lenny2. Since then GRUB has problems booting properly. Initial boot throws up "Error 13: Invalid or unsupported executable" and stops. And here things go weird: the menu list contains a manual entry to an installation on another disk which is no longer there. Launching that entry results in a "disk doesn't exist" error as expected. But after that retrying the real entry now launches the kernel perfectly... Similarly: using the GRUB shell to edit the root stanza to point to a non-existing disk, attempting to boot, then restoring it to the original entry makes the kernel boot. Something in there is making GRUB change its mind on where to look for the kernel. The reason I suspect the softRAID is that the menu list displayed on boot does not match the one in /boot/grub. I found one possibly related Debian bug (526525) that hints at an AMD64 specific problem, so for those with a similar setup, watch out...
Post #313,154 by static 8/27/09 9:26:14 AM Reply	Makes me glad I never RAIDed my work box. All the developers at work got massive box upgrades a few months ago: standard box is now a dual dual-core AMD 64-bit thing with 8Gb of RAM, 2x300Gb HDDs and Vista 64-bit. The .NET and Java developers in particular wanted all the extra RAM and extra cores for their compiles. But 'cause I develop in PHP, I put Ubuntu on it instead, and a helpdesk tech helped me put Vista into a virtual machine for Outlook. I was a bit surprised the Ubuntu CD didn't offer any LVM option, so I installed it without, but realized later that it would have only been on the Alternate CD. Wade. Q:Is it proper to eat cheeseburgers with your fingers? A:No, the fingers should be eaten separately.
Post #313,170 by folkert 8/27/09 3:52:27 PM Reply	Re: FYI - SoftRAID weirdness after kernel 2.6.26-2-amd64 upg Similar problems bit me recently on Sid. Except it was for udev/hal and removal of sysfs support
Post #313,617 by scoenye 9/5/09 8:28:53 PM Reply	The kernel was innocent. Lessons learned... Always check /proc/mdstat before rebooting after a kernel upgrade... This is indeed something that could use a big red warning flag of sorts. What happened is that one of the two disks had been kicked out of the array for whatever reason (but not a disk fault). This does show in dmesg, but is buried in the pile. The BIOS reported a healthy array, so I wasn't paying much attention to it. As a consequence, the upgrade was installed on only one disk. The boot sectors are not part of the array and GRUB was using the sector on the ejected disk to fire up. On reboot, this caused GRUB to look in the wrong place for the kernel image. I still don't quite get why editing the boot stanza caused it to jump to the good disk.

Welcome to IWETHEY!