Booting with LVM2 is much too slow

[Ben Steeves]

On Mon, 2004-03-22 at 20:47, Neal Becker wrote:
> Now that I'm testing with lvm2, there is one thing that must be fixed before
> release.  During boot, it says, "activating logical volumes", or something
> like that (not logged to dmesg), and then ***sits there for a good 5
> minutes with no user feedback that it is even alive***.
> 
> OK, so exactly why did I install a journalling filesystem?
> 
> Seriously, I hope it's only like this due to some debugging and it won't act
> this way when released.

This is what Alexandre Oliva wrote when I posted a similar question:

On Feb 13, 2004, Ben Steeves <bcs at metacon.ca> wrote:

> I just installed a FC2 Test 1 system and applied all of the days
> patches.  When the machine boots, it sits at teh "Setting up Logical
> Volume Management" phase for a very long time (about 5 minutes). 
After
> this delay, the message "setlocale failed" is printed and boot
continues
> normally.  The filesystems all appear to work correctly.

> Is anyone else seeing odd LVM behaviour?

lvm vgscan scans pretty much every block device it can get its hands
on.  You may filter out stuff that doesn't matter for you using a
filter directive in /etc/lvm/lvm.conf.  Here's what I use for a
machine whose physical volumes are *all* raid devices:

devices {
    filter = [ "a|/dev/md.*|", "r|.*|" ]
}

this is for another box in which the `r'(emoved) devices are RAID
members:

devices {
    filter = [ "r|/dev/hd[acegimos][5689]|",
"r|/dev/hd[acegimos]1[012]|", "a|.*|" ]
}

skipping raid members is a big problem that I wish lvm could take care
of by itself, otherwise anaconda will have to do it by itself.  If you
don't skip raid members, lvm vgscan will simply bail out after finding
some ``duplicate'' physical volumes, that turn out to be just the RAID
1 members and the RAID device itself.  Ugh!

Anyhow, limiting the scanning to only the devices that matter for you
(with "a|regexp|", skipping everything else) will help speed up the
boot very significantly.
-- 
Ben Steeves                     _                    bcs at metacon.ca
 The ASCII ribbon campaign     ( )               ben.steeves at unb.ca
   against HTML e-mail          X                GPG ID: 0xB3EBF1D9
http://www.metacon.ca/ascii    / \     Yahoo Messenger: ben_steeves