Re: nonoptimized kernel compilation (an idea)

Ton Hospel (thospel@mail.dma.be)
27 Jun 1998 17:19:29 GMT

In article <Pine.HPP.3.96.980625171020.11617D-100000@ixion.honeywell.com>,
Shawn Leas <sleas@ixion.honeywell.com> writes:
> On Thu, 25 Jun 1998, Chris Siebenmann wrote:
>
>> People (including me) want to do operations such as growing a disk (or
>> a disk set) in the middle of a filesystem. They don't want to add parts
>
> You cant, unless you do it in the LVM and make it transparent to the FS.
> Otherwise, as you mentioned, it becomes too complex.

I too would like to be able to replace disks by bigger ones. on the LVM's
I've used, replacing disks by the next generation (typically bigger) is in
fact the operation I do most, once I thought of and implemented the initial
setup.

What if we give up that mirroring has to match PV boundaries,
then you could do:

old: <-----2M-----><-----2M-----><-----2M----->

Now I buy a 3M disk and want to replace the middle.I do:

<-----2M-----><-----2M-----><-----2M----->
<---------3M-------->

After the second 2M disk is completely mirrored, I remove it:

<-----2M-----> <-----2M----->
<---------3M-------->

And now add it to the end:

<-----2M-----> <-----2M----->
<---------3M--------><-----2M----->

I wait 'till the old last 2M disk is completely mirrored, so I can remove it:

<-----2M----->
<---------3M--------><-----2M----->

or collapsed:

<-----2M-----><---------3M--------><-----2M----->

where the last 1M is now unused and we are free to grow the FS into it.

This would seem to allow a probably good enough form of
adding bigger new disks, without the FS ever seeing it.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu