[conspire] request verification of plans to partition and format external hard drive

Darlene Wallach freepalestin at dslextreme.com
Fri Jul 20 16:10:24 PDT 2007

Rick Moen wrote:
> Quoting Darlene Wallach (freepalestin at dslextreme.com):
>>I have a 160G external firewire hard drive.
> Lucky you.

Thanks to my former next door neighbor who moved out and
needed to get rid of a lot of hardware.

>>1. Should I make 2 partitions?
>>    - one for data - all the rest for data
>>    - one for swap - how much swap?
>>2. Can I make one big data partition?
> That's certainly the easiest option, and is perfectly OK as a starting
> point pending you developing your own strong opinions about partitioning
> practices and going around advocating them.[1]  ;->
>>3. Does it matter which I partition first? I was
>>    planning on putting data on 1st with swap at
>>    the end.
> There might be a slight performance advantage in putting the swap near
> the _front_ of the drive, in such a two-partition configuration -- on
> grounds of minimising average seek distance (and thus seek time).  This
> is one of the many, many points of dispute sysadmins happily argue
> about, for hours on end.

Thank you, I will put swap on first and give swap 2G,
following what I read in your footnote the
nix-partitioning.html by Karsten. Since I'm not using
the external drive as a server, I'm using Karsten's

>>4. Since I may use this external drive for data, I plan
>>    on using ext3. Is that a good plan?
> Sure.  ext3 is a fine general-purpose filesystem for Linux.
>>5. Should I use the "s" option in fdisk to create a new
>>    empty Sun disklabel vs "n" for add a new partition?
> For use on Linux, it's better to create the IBM/Microsoft-style "PC"
> partition table that results from using the "n" command and avoiding
> creating a Sun (or BSD) disklabel.  
> On an earlier incarnation of linuxmafia.com, I _accidentally_ had a BSD
> disklabel rather than a "PC" partition table on the 2nd hard drive for a
> couple of years, because that drive had recently had FreeBSD on it.
> Linux seemed to _mostly_ deal with that transparently, though there were
> a few operational oddnesses that I thought _might_ have traced back to
> the non-standard partition table, so I eliminated that variable when I
> rebuilt the box.

Thank you, "n" it is.

> [1] http://shearer.org/Linux_Server_Partitioning  <= Dan Shearer
>     http://linuxmafia.com/faq/Admin/nix-partitioning.html <= Karsten M. Self

When I select the +sizeM for swap of 2G, is this
the correct size: 1024M * 2 = +2048M
+2048M for swap

Then I should select the "t" option for the swap

Then I do I still need to?
mkswap -c /dev/sda1 2097152
swapon /dev/sda1

When I select the +sizeM for data, is this the
correct size for data? 169G - 2G = 167G
1024M * 167 = +171008M
+171008M for the data

Then I should use mkfs.ext3
I read in man fdisk that using "-c" twice a slower,
read-write test is used instead of a fast read-only test.
mkfs.ext3 -c -c /dev/sda2 2097152

Thank you,

Darlene Wallach

More information about the conspire mailing list