[GRLUG] Active Partitions

Greg Folkert greg at gregfolkert.net
Wed Dec 5 11:50:58 EST 2007


On Wed, 2007-12-05 at 10:50 -0500, Justin Denick wrote:
> On Dec 5, 2007 10:26 AM, Greg Folkert <greg at gregfolkert.net> wrote:
> >
> > On Tue, 2007-12-04 at 14:50 -0500, Justin Denick wrote:
> > [snip]
> > > FWIW, I did not try GRUB. Grub is garbage.
> > [snip]
> >
> > So, exactly why do you think this?
> >
> > I have a similar opinion of all boot loaders, not just GRUB, but LILO,
> > MILO, SILO and a boot loader from Microsoft, IBM's boot loaders for OS2
> > and such. This along with other boot managers that are out there. They
> > all SUCK.
> >
> > Care to enlighten me?
> 
> LILO has been my bootloader since I began using linux. I am a creature of habit.
> I used GRUB once, it failed me, so I have not gone back.
> Since I don't dual boot anything, and I only keep one kernel, booting
> is just a means
> to an end. I suppose if I had began using GRUB, and LILO had failed
> me, I would think
> differently.

I started with something else, I can't remember what it was called.

Perhaps it was LILO, but it wasn't called that when SLS Linux first came
out.

> > FWIW, LILO needs to know WHERE EXACTLY (C,H,S) the booting image/kernel
> > is in order to load it. *THAT* is why you have to "re-run" LILO every
> > time you install a new kernel.
> Not a big deal, since running make install after compiling your kernel
> now prompts you.
> I'm not sure when that started, I think it happened after 2.6.10 or 2.6.12.

Hmmm, didn't know that.

Until recently (few years) LILO couldn't go past the 1GB boundary on
disk *RELIABLY*. Sure it has been able to for a long time, but you had
to ensure you really has the *right* version of LILO.

> > Also, GRUB doesn't need to know exactly where the booting image is, but
> > it does have to be able to *read* the file-system upon which the booting
> > kernel/image, or chain-load it from things like Windows or OS2.
> >
> > Personally, I use GRUB only because it is more reliable for remote
> > machines, compared to LILO.
> 
> A piece of software that forgives operator error, should not be deemed more
> reliable than one that doesn't.

Nope, we are talking about NOT operator error, but DISTRO PROVIDER
ERROR. Specifically RedHat Enterprise Linux v3 update 9 and v4 update 5.

These machines had been upgraded since forever. Reboots are always
fraught with problems when using LILO. Even when properly setup, there
is a good chance that LILO will not properly detect the right setup.

We are talking machines with 30+ disks and 2-5 drive controllers, about
8 kernels (smp and up and revisions if problems occur with the new
kernel) installed at any one time.

One feature GRUB has is command line booting, in other words, even *IF*
someone blows away the menu.lst, you can still boot the machine FROM the
GRUB prompt, assuming you know how to do this.

LILO failure == booting a recovery disk and hopefully having the
recovery disk evaluate the numerous disk in the same order so that LILO
writes the (apparent) MBR properly.

> > Of all the machines I've had not reboot properly, *recently*, all have
> > been LILO.
> 
> Was it LILO's fault?

Actually it was. 30+ disks, some as mirrors of others (especially the
booting partition) causes LILO to get confused occasionally. It works
99% of the time, but fails 1%. It fails at the least opportune time, as
has been my luck. I still have about 70% of legacy machines using LILO,
so don't think I am "fringe" case.

        I have 2 new machines using LILO, as we have a requirement to
        use XFS only on any filesystems... its a "third party" app that
        the vendor has "strict requirements"... and they have installed
        new kernels and lotsa other stuff rebooted that machine... and
        have had to pay to have someone "work" on the machine to get it
        to boot. And yes, its been operator error in this case.
        
        FYI on that, they recently went from supporting ONLY Solaris 9
        on SPARC hardware, to RHEL v4 update 3 setup this way. These
        machines are now, 2 quad core processors, blahblahblah machine
        with 3.5GB of memory (running ia32 kernels). I am tempted to
        migrate them to VMs, as the machines *NEVER* see any load
        average above 0.2. I'll bet I could do it without them even
        knowing. Until its is too late for them.

FWIW, GRUB never needs to be re-run. You just change the config file. If
you make a mistake in that... you have the "edit" function available to
fix the problem (at boot) OR you can go into command line and manually
load the kernel yourself.

So, given that we are all prone to mistakes in any case, which one would
you use, based on using Linux for your CORE business model and
application serving?

> > Same machines, once GRUB was installed and setup properly (being trivial
> > to do), have not had a a single problem rebooting since.

-- 
Greg Folkert <greg at gregfolkert.net>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://shinobu.grlug.org/pipermail/grlug/attachments/20071205/aada4199/attachment.pgp 


More information about the grlug mailing list