Checking huge filesystems [was: Re: Resizing /var (maybe off topic)]

[Giorgos Keramidas]

On 2005-06-12 14:41, Wojciech Puchar <wojtek at tensor.3miasto.net> wrote:
>
> i'm using different unices for 7 years and excluding few cases i never
> made other partitioning scheme than 2 partitions: swap and /
>
> i have no problems like "there's out of space in partition x while
> plenty of y".  it's far easier to do backups too (single dump).
[...]
> fsck speed is same in checking one partition or many smaller. and it
> doesn't matter at all as FreeBSD (and NetBSD) doesn't crash every few
> hours like windows

While you have made a few good and valid points, this last one is not
really true.

FWIW,

Checking a filesystem involves allocating enough memory to reconstruct
the inode and data block allocation bitmaps, when something is wrong,
along with other meta-information.  This information is kept in data
structures that are traversed many times.

It may be possible to check 20 filesystems of 100 GB each a lot faster
than a single 2000 GB filesystem, if the fsck process that is supposed
to check the latter runs out of memory because of the filesystem size.