e2fsprogs, com_err, and ss blockers

jserink · Guru Joined: 30 Jan 2004 Posts: 559

Ok, rebooted fine, sweet.
Back up and running.

:)
John

krotuss · Tux's lil' helper Joined: 01 Aug 2008 Posts: 86

What is ETA on fix?

Moriah · Posted: Mon Nov 03, 2008 3:24 pm Post subject:

Well my box that boots from a raid1 under lvm refuses to even try to boot. All I can think of is that when e2fsprogs-lib got clobbered, that somehow, something that used ext2, which is what my boot partition uses, something like that ran amuck and maybe clobbered a partition table or something.

When I try to boot the box, I do not even get to the grub screen where you get to choose what image to boot. The BIOS complains that the drive is unbootable.

_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

marcus0263 · Posted: Mon Nov 03, 2008 5:42 pm Post subject:

Until they get this fixed I just added in my /etc/portage/package.mask

>=sys-libs/e2fsprogs-libs-1.41.0
>=sys-fs/e2fsprogs-1.41.0
>=app-crypt/mit-krb5-1.6.3-r4
_________________
Reject The Herd
"Where Thought Crime is Committed"
www.rejecttheherd.net
Gentoo Powered

Stupendoussteve · n00b Joined: 06 Sep 2005 Posts: 72 Location: NE, USA

Moriah · Posted: Tue Nov 04, 2008 2:47 am Post subject:

Well, I got the 500 GB drives, installed them, partitioned them with a 50 MB boot partition, and the rest of the drive devoted to the root filesystem. Used mdadm to create raid1's on the 50 MB boot partitions, and on the nearly 500 GB root partitions. Waiting for the big raid1 to resync.

Next step will be to put ext2 filesystem on the 50 MB /dev/md0, and use LVM2 to manage the big partition, allocating 3 GB to swap, 50 gb for spare space for snapshots during backups, and the rest for the XFS filesystem that will be the root filesystem after the pivotroot during boot.

Then its on the the normal Gentoo install and a kernel build. Finally a grub installation, including writing the MBR.

Then the first boot. (pray, pray, pray...)

Then install all the other stuff this server needs, like apache, php, squirrellmail, mysql, sendmail, etc. etc. etc. :roll:

Then I have to restore all the files from the backup server unless they already exist on the new installation.

I also am just starting to work on the clobbered dhcp/samba/cups server machine, then I have to get the NIDS box running again, then I finally get to work thru this tricky installation on the other 8 Gentoo boxes here.
_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

ade05fr · Apprentice Joined: 18 Sep 2006 Posts: 182

Up again

Moriah · Posted: Tue Nov 04, 2008 8:44 pm Post subject:

Well, I just went thru the gentoo installation procedure, using raid1 and lvm for the root filesystem, and booting from that same pair of drives. There are really 2 raid0's here: a 50 MB etx2 filesystem on /dev/md0 as /boot, and the rest of the disks as /dev/md1 with lvm on top of that, and a swap partition and a root filesystem partition running xfs.

I did the whole install from scratch thing, including emerging grub and running it to put the good stuff in the MBR etc. on both drives, AND THE STINKER STILL WON'T BOOT!!!

I get the same symptoms as before: not even an indication that it is so much as trying to run grub -- only a message from the BIOS that the drive is not bootable. :evil:

So now I am dropping back to really simple, and doing an ext2 install on only one drive. Both /boot and / are nice simple ext2, and there is only 1 drive in the configuration. We shall see if it will boot.

I suspect it will. I am starting to suspect that my memory on how to set up grub to boot from a pair of drives as raid1 mirror might be a bit faulty...

_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

Moriah · Posted: Tue Nov 04, 2008 11:43 pm Post subject:

Well, it DIDN'T boot!

Then I noticed that the instructions for grub.conf now show /boot/kernel... instead of just /kernel... like it always used to be. Has somebody changed the way grub wants grub.conf set up?

Well, I tried it both ways, and even re-ran grub after teach attempt, and still to no avail.

So now I am beginning to suspect that some weird hardware problem has arisen that keeps it from booting, even though I can read and write the drives from the live CD just fine, and the live CD boots fine too.

So to answer that question, Lord forgive me, I am installing a copy of XP to see if that will boot.

I will post back after this test... :roll:

_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

Moriah · Posted: Wed Nov 05, 2008 12:33 am Post subject:

Well, XP didn't boot either!

Yes, in this case, that's a GOOD thing. It means I have a problem that is not a botch job on my part of installing gentoo.

So I took a good close look at the BIOS settings. I tried many combinations, but nothing worked.

So I tried the "reset to defaults" BIOS choice.

Voila! XP booted!

So it would seem to me that something ran amuck due to the com_err/ss fiasco, and clobbered something in my NVRAM that holds the BIOS settings.

Has anybody else seen anything like this? :?:

Well, since I think I have solved the mystery, its back to the gentoo install. Given what I have been thru so far, I will do the simple ext2 only install first -- to prove that it will boot -- and them move on to the boot from an XFS filesystem under LVM under RAID1. If all that works, then I get to rebuild the rest of my server.

What a nightmare! Yes, it does remind me of the expat disaster, only maybe even worse! :evil:

(And to make matters even worse, the US presidential election returns are starting to come in on the radio streaming in over the internet...

)
_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

hitachi · Guru Joined: 20 Feb 2006 Posts: 461

@Moriah: What's the relation to "e2fsprogs, com_err, and ss blockers"?
_________________
Athlon 64 X2 3800+
Micro-ATX-Mainboard in einer OEM-Version von Asus mit Nvidias nForce4 (A8NE-FM/S).
PEG-Grafikkarte von Leadtek mit Nvidia GeForce 6600
3 GB RAM
350-Watt-Netzteil (von Fortron/Source)
4 x 500GB S-ATA HDD (raid1 und raid5)

Moriah · Posted: Thu Nov 06, 2008 1:46 am Post subject:

My understanding is that e2fsprogs calles routines in com_err and also in ss. com_err is "common error handler", and ss is "subsystem". Lousy names, and when you unmerge e2fsprogs, it unmerges these 2 other libs at the same time.

But the poor coupling is that *MANY* other things also call com_err. Things that include wget, which ought not have anything to do with the e2fs filesystem, but they just knew that com_err would be there, so they called it like it was hardwired into the cpu or something.

Because of this terrible coupling situation, together with the undeclared dependencies that result from these other programs calling stuff in com_err, when you unmerge e2fsprogs to get rid of the block, you also send wget south for the winter, which, of course, is for the birds! :evil:

What seems to be working for me so far is to boot the broken system from a live CD and mount your old root filesystem on /mnt/gentoo. Use your alt-F1 and alt-F2 to switch between virtual consoles, because you are going to chroot on one of them, and keep the other in the outside environment.

Now, you can go into the chroot shell and do an emerge -pv --update --deep --newuse world and see if the block is still there. If it is, then you haven't broken that box yet, but if the block is gone, it means your wget is trashed.

Id your wget is trashed, go to the non-chroot shell and cd to /mnt/gentoo/usr/portage/distfiles and do a wget com_err ss to manually pre-fetch the files for com_err and ss. Next, go to the chroot shell and emerge com_err and ss. Since the files are pre-fetched, you will not call the messed-up wget. When the emerge is done, your wget will be resurected, having risen from the dead. Now you can pre-fetch everything else with an emerge --fetchonly --update --deep --newuse world.

When the mega-pre-fetch is done, you can once again unmerge com_err and ss to remove the block, and then do you emerge --update --deep --newuse world to finish the recovery and the update of your system at the same time.

Now exit the chroot shell, unmount /mnt/gentoo, and reboot the box after removing or disabling the live CD.

Everything should work OK at this point.

_________________
The MyWord KJV Bible tool is at http://www.elilabs.com/~myword

Foghorn Leghorn is a Warner Bros. cartoon character.

Stupendoussteve · n00b Joined: 06 Sep 2005 Posts: 72 Location: NE, USA

Not everyone ends up with a trashed wget from removing those, only if they are using mit-krb5 and it depends on those. If they are removed then it breaks, and breaks something else (openssl I believe) which breaks wget. The fix was pretty straightforward, especially since the fixed mit-krb5 made it to stable.

As for the grub.conf, pointing to the kernel is done based on the grub root. If your grub root is (hd0,0) and /dev/hda1 happens to be your /boot partition, then you would only need to point to /kernel. If you are using a / partition then you point to /boot/kernel.

By the way, the better way to fix it if you've hosed your system but booting from a cd is to download the new versions into distfiles, not com_err and ss. Upgrade the packages, fix the block, and get on with the rest of life. No need to reinstall the old packages just to do the same thing.
_________________
Gentoo - FreeBSD - Debian - Arch Linux

mariourk · Posted: Fri Nov 28, 2008 10:43 am Post subject:

justincataldo · Guru Joined: 15 Jun 2005 Posts: 376 Location: Australia

This worked for me. Thanks!

jk3us · Posted: Tue Dec 30, 2008 8:04 pm Post subject:

A note if you use "sudo emerge" if you unmerge ss and com_err, which looks like a preferred solution, you will not be able to run sudo until you get e2fsprogs/-libs installed. 'su-' still works to get you a root shell, but I almost couldn't remember my root password.

The point being... run those steps as root, from the root prompt, not via sudo.

Anyway, I finally have this problem fixed (had just package.mask to hide it for a while)