[ale] eth numbering change

[Jim Kinney]

On Mon, 2017-02-13 at 16:18 -0500, Phil Turmel wrote:
> On 02/12/2017 12:42 PM, Steve Litt wrote:
> > On Thu, 9 Feb 2017 13:38:07 -0800
> > In addition to what Alex writes in the preceding paragraph, some
> > people
> > prefer not to have initramfs at all.
> 
> Yup.  Embedded guys really don't want anything extra.  And they
> customize their kernels to give them exactly what they want.

HPC has similar end needs. Even a module hook takes RAM. When the
hardware is not going to change over it's lifespan, a custom kernel
with no extra bells or whistles is nice.
> > The purpose of initramfs is ONLY to get the root partition readable
> > enough to process /etc (let's forget about crazycases where /etc is
> > a
> > mountpoint), after which the boot process can do what it's supposed
> > to
> > do: Pull itself up by the bootstraps. The initramfs' init script
> > should
> > have been an easy, tiny shellscript in most cases.
> 
> People all over the world want faster boot.  The time spent
> initializing
> just the devices needed to get to the boot partition and then mount
> its
> filesystem can cover a great deal of parallel probing and
> discovery.  In
> some cases, all of it.  People ask distros for faster boot times,
> distros respond.  So what if there is substantial complexity needed
> to
> deliver what their customers want.
> 

If we could only get the HARDWARE side of the boot process faster! I've
got some very new, very modern server gear that from power on to Linux
boot start is measured in MINUTES (I'm looking at you IBM!). Why oh why
can the BIOS store the IDs of the device tree and do a fast "give me
your ID scan", ok nothing changed so send back an init command to
everything to run in parallel.
> > My personal observation tells me that with kernels ancient and
> > modern
> > (I'm using 4.9.x right now) and everything inbetween, disk device
> > names
> > are stable between boots. So personally, I might consider using the
> > device name in order to perhaps compile ext4 drivers into my kernel
> > and
> > getting rid of the entire initramfs.
> 
> If you only have one SATA controller, one connected drive, and you
> defer
> USB module loading, you can be pretty confident that your drive name
> will be consistent.  Or you customize your kernel to do exactly what
> you
> want.  If you're willing to do that, you need no-one's advice.

When there's 2 drive controllers, 2 RAID cards, 2 nics on board and 2
10G cards and a dual infinniband card, walking that PCI bus takes
FOREVER.
> Everyone *not* willing to customize, but using a modern kernel, needs
> to
> use an initramfs and proper partition IDs.

I keep swearing (more than usual) I'm gonna build by blasted tool
called "shoehorn" to help squeeze into a tight boot. The basic idea is
to PXE boot a generic kernel with EVERYTHING as modules, probe all
hardware, report back to a master server the list of found hardware, it
assembles a new kernel with nothing but what's needed or points to an
existing kernel, set's up a perm PXE boot environment based on MACs and
then provides the slimmest kernel and initramfs possible.
> Phil
> 
> _______________________________________________
> Ale mailing list
> Ale at ale.org
> http://mail.ale.org/mailman/listinfo/ale
> See JOBS, ANNOUNCE and SCHOOLS lists at
> http://mail.ale.org/mailman/listinfo
-- 
James P. Kinney III

Every time you stop a school, you will have to build a jail. What you
gain at one end you lose at the other. It's like feeding a dog on his
own tail. It won't fatten the dog.
- Speech 11/23/1900 Mark Twain

http://heretothereideas.blogspot.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ale.org/pipermail/ale/attachments/20170213/a8ee51fa/attachment.html>