Re: your mail

[Hendrik Boom <hendrik@topoi.pooq.com>]

On Wed, Aug 07, 2013 at 07:36:28PM -0700, Alan Hutchinson wrote:
> Hi there My name is Alan Hutchinson and I would like to install Debian on a
> raid 0 system that I have at home , on my home network that I created, I
> have bean reading up on some of the things that I will need to complete a
> successful install, for example the raid manager (MDADM) and you also
> mentioned 
> DebianInstaller does not include a driver for my xyz network card, but I
> have (tarball) driver sources. What are the steps for building that driver?

My home server has a RAID for most of its storage.  It's a software RAID.  
WHen I bought the controller card for my machine (so I could get more SATA 
drives on it) it turned out to be a RAID controller.   I completely ignored 
its RAID capabilities,  opting to use Debian's excellent md support instead.

I created large primary partitions on my IDE hard drives to contain one RAID, 
and large partitions on my 3-terabyte drives for the second RAID.  Yes, 
that's right, two independent RAIDS -- the IDE drives for legacy stuff, the 
new 3T drives for newer (ald larger) stuff.  My long-term plans are for the 
IDE drives to be replaced by still bigger drives in a year or two when they 
become necessary.

And within those RAIDs I used LVM for flexibility in allocating partitions 
of flexible size.  

To make everything work smoothly, hpwever, it helped to
  (a) use gdisk instead of fdisk onthe new huge drives. fdisk doesn't work 
anh more in 3T drives -- they're too big.
  (b) make sure that there was a working boot partition outside of the 
RAID. Some bootloaders had problems with boot partitions on LVM on RAID.
  (c) made sure that the first physical partition on the drive contianing 
/boot starte *after* the first cylinder.  Apparently LILO now needed more 
space than it used to, and it needs to invade the first cylinder.
  (d) I use the badblocks program to do a full-surface destructive read-write 
test of every disk drive I buy.  Yes, it can take longer than a week, so I 
buy them one at a time, even if I plan to use them in pairs.  But about 20% 
of them fail the test, and are immediately exchanged.  I no longer worry that 
I'm being excessively paranoid.

I'm now testing my second USB backup drive.  My plan it to keep at least one 
backup drive off-site at all times.  I use rdiff-backup; it keeps history.
 
Also, I made sure I could boot from either grub2 in the usual manner, 
or from LILO on a floppy.  I test them regularly to make sure they 
both work.  If either ever fails during an upgrade, I still have the other. 

Finally, I'm doing the whole thing using wheezy.  It seems to be the first 
Debian release that fully supports disk hardware write blocks -- even 
through LVM -- so that file systems like ext3 and ext4 have a reasonable 
chance of processing their journals properly even in the presence of 
unexpected power failures.

-- hendrik

> 
> A: It's possible to do this, but we do not yet have good end-user docs
> explaining how to do so. We're waiting for an end user to accomplish it and
> write them from experience. We're glad to help you through the process, I
> have a brand new D-LINK 2320 that I would like  to install with  Debian
> installation, I have not tried to do an "LSPCI-V" on the raid hardware as
> yet, I can do that with PUREDYNE, and send you the data ,  B.F.N  Alan
> Hutchinson .

-- hendrik