Andy,
I have similar requirements and have went in search of how to do this through
kickstart/anaconda. For RHEL 5 I ended up using vmware templates which was
something I had hoped to avoid so it's good to hear that hopefully with RHEL 6
I might be able to get away from this at least with %pre functionality. My Red
Hat technical contact was supposed to file through internal channels an RFE on
this as well which I need to check on. Do you mind sharing the BZ# you
submitted so I can reference this as well. I haven't had a chance to see if the
rescue environment has caught up with this practice or not.
The reason I had looked at the whole disk model was for our virtualized
environment as well on VMware. We usually do disk/file system
expansions dynamically. With RHEL 5 on VMware I attempted to get away from our
historical model of addind a new LUN, partition, and then add to LVM, etc.
There was also some performance recommendations from VMware to keep the number
of virtual disks low per VMFS volume and other VM sizing considerations (# of
LUN's per virtual storage controller).
For my VM's I am essentially creating a single 512m partitioned disk for /boot,
and then a large unpartitioned disk for the LVM PV which hosted everything
else. Whenever I needed to increase space I would resize the second LUN in
VMware, poke the correct sysfs items, & then run a pvresize, etc.
I'm assuming I can probably do the samething with partitions by resizing the
partition (assuming it's at the end of the
disk) that hosts the PV but I didn't have any extensive experience
with resizing partitions. I have had past issues with changes to the partition
table requiring reboots so I shyed away from really exploring this too deeply
based on the amount of time I had. Altering the partition table on a live
system also gives me and other system admins here the creeps :-).
Chad
________________________________
From: "Speagle, Andy" <andy.spea...@wichita.edu>
To: Discussion list about Kickstart <kickstart-list@redhat.com>
Sent: Tuesday, May 10, 2011 12:36 PM
Subject: RE: Use entire disk as PV.
> Speagle, Andy wrote: > Sent: 10 May 2011 13:53
> > > Did you use a double dash for “--preexisting”? I didn’t notice as I
> > was typing it that Outlook had
> > > “helpfully” changed the two dashes to a single emdash. Did the
> > pvcreate command actually work: does that
> > > disk now have a full-disk PV on it? In the pvcreate documentation,
> > > I
> > just came across
> >
> > Aye, I had used the --useexisting option with the double-hyphen.
> >
> > > Here’s something else to try:
> > >
> > > echo "part pv.01 --onpart sdb --grow" >> /tmp/custom-part echo
> > > "volgroup rootvg /dev/sdb pv.01" >> /tmp/custom-part
> > >
> > > If kickstart refuses to create a partition on sdb, you will have to
> > do something like
> >
>
> > > lvm pvcreate /dev/sdb
> > > lvm vgcreate rootvg /dev/sdb
> > > echo "logvol / --fstype ext3 --name=root --vgname=rootvg --size=4096
> > --useexisting" >> /tmp/custom-part
> >
> > This seems to have worked for the RHEL 6 installer. RHEL 5.5 still
> > barfs on this saying that no VG exists with the name "rootvg" ...
> >
> > I have filed a BZ with this as a feature request. I hope this gets
> > added at some point. It should be fairly trivial to do.
>
> Is it any better in RHEL 5.6?
> Were you able to tell if the lvm commands in the %pre script actually
> worked? Try them manually from a kickstart shell; that may give a clue
> where things are going wrong. If the volgroup is being created, maybe
> kickstart in RHEL 5.5 isn't re-reading the disk information. Maybe
>
there's some way to get it to do that, or at worst get your %pre script to
> detect whether the disk has volgroup information. If not, create it and
> reboot; else write /tmp/custom-part to create the logical volumes and
> continue the install.
RHEL 5.6 suffers the same fate. I received a response on my BZ for this
feature request that native support for PVs on entire disk will be included for
anaconda somewhere in RHEL 6 ... but that it will not be back-ported to any
RHEL 5 anaconda.
All of the LVM commands are working just duckily. Anaconda just refuses to
handle this situation. The error is quite explicit that it's an issue parsing
the kickstart file. Anaconda is expecting a "volgroup" definition and won't
continue without one.
-Andy
_______________________________________________
Kickstart-list mailing list
Kickstart-list@redhat.com
https://www.redhat.com/mailman/listinfo/kickstart-list
_______________________________________________
Kickstart-list mailing list
Kickstart-list@redhat.com
https://www.redhat.com/mailman/listinfo/kickstart-list
