[ILUG] dying hard disc - upgrade questions?

[Brian Foster]

 hi!

   the hard disc in my home Linux system is dying....  ;-(

   apologies for the length of this posting, it contains
  a bit of background, some speculations and questions,
  and perhaps a lot of rambling ....

   the drive's sounding like a washing machine, and
  the SMART data suggests it is starting to register
  the odd error or two.  it has also hung at least twice
  (i.e., access LED goes on and never off, and the system
  eventually freezes), and makes terrible noises when
  spinning up after a power-on (think of howling cats
  being thrown into a blender whilst the more militant
  sort of animal rightist is blowing things up and you'll
  only have ½ of it).

   consequentially, I will be replacing it soon, and in the
  process --- to see how many things go wrong at once ---
  also installing a recent distro (SuSE 9.1 (kernel 2.6)),
  as my somewhat old SuSE 7.3 workhorse is starting to show
  its age.  (still as reliable as the better sort of rock,
  however!)

   the paranoid presumption I am using is that once the
  system is powered down for any significant period of
  time, the current hard drive will *never* spin up again.
  hence, the general plan is:

  ① to sacrifice a few of those cats to the local gods
    (and pray there isn't a power failure);

  ② backup, backup, and backup yet again (to CDR);

  ③ install the new drive (maybe keeping the old drive in
    the system, maybe not, depends on issues like physical
    access to the drive bays, jumpering, &tc);

  ④ install the new system; and

  ⑤ restore as appropriate (which is when having the
    current drive, if it is still working, also still in
    the system would be very useful --- its quicker to copy
    from the old drive than find the correct CDR backup).

   I've been going through the relevant LDP HowTo's for
  hints, etc., about hard discs.  first problem is I'm
  not completely certain exactly what is in my machine;
  I know it is, broadly speaking, ATA, but I don't know
  if it is the original ATA(-1), ATA-2, or ATAPI (I hope
  it is the later, then some of my questions go away!).
  and I vaguely seem to recall the controller is on the
  motherboard (and not a PCI board).

   the issue here is that, AFAIK, ATA-1/-2 have a limit
  of c.137Gb, suggesting my replacement must be ≤ 120Gb,
  which is Ok, albeit I was thinking of c.200Gb; the only
  firm size requirement is ≥ 80Gb.  (the dying 40Gb will
  also start to present space-management problems later
  this year.)  OTOH, ATAPI supports drives into the Tb(?)
  range, and so is not an issue.

  QUESTION:  how can I work out if my controller et.al.
            is ATA-1, ATA-2, ATAPI, or something else?

   also, some of the replacement drives have “cache‟
  (no further details known).  thinking about it,
  I don't see an issue with WT (write-through) cache,
  but WB (write-back) caches can cause data integrity
  problems (e.g., on power failure), albeit if the
  driver/ATA/IDE/cache(/BIOS?) supports flush and barrier
  operations, then it is perhaps Ok.

  QUESTION:  should I steer clear of the drives with
            “cache‟, or are they generally Ok?

   on the matter of physically installing the new drive,
  I am aware it will need to be jumpered appropriately,
  albeit I'm not quite sure what “appropriate‟ means
  here .... presumably as the master(?) if it is the only
  drive; if the old 40Gb is still present (which I thus
  assume is the master), as the slave(?) --- but then can
  I boot _exclusively_ from the slave without the master
  being involved (or even present?)?

  ( related to the above confusion is that it would be
   bloody useful to still be able to boot the old drive
   (assuming it is still working), at least until I get
   Internet access working with the new system; and so,
   at the least, if it is no longer the master(?) drive,
   then any hardwired names --- and I know there are
   some --- like /dev/hda<N> in the existing SuSE 7.3
   system won't be correct. )

   one thing I would like to do given a bigger drive
  is to set aside space for at least one, maybe two,
  alternative distros to play with (e.g., some of those
  that appear every now and then in the LXF cover DVD);
  this includes other Linux distros and also both Plan9
  and the various BSDs.

   ergo, it would be wise to keep at least one, maybe two,
  primary partitions “free‟ ... which means I would
  have to use extended partition(s), which is something
  new to me.  the general arrangement would then be:

    /boot  partition 1, within 1st 1024 cylinders, either
           LILO or GRUB (depends on what's in SuSE 9.1).

    x      partition 2, physically follows /boot,
           “scratch‟ and/or alternative distro.

    y      partition 3, at the other end of disc, also
           “scratch"/alternative distro.

    extended partition 4, usual collection of /, swap,
           /home, and so on, dedicated to the new SuSE
           9.1 installation.

  ( QUESTION:  anyone happen to know if SuSE 9.1 easily
              supports a read-only root? )

   the obvious alternative (but not mutually exclusive)
  is to use a 2nd drive, which has the added attraction
  that it could be easily added/configured later; and
  could be used in a backup strategy....

   I've read conflicting (albeit not contradictory)
  explanations of extended partitions.  the combined
  ghist seems to be they form a chain.  the contents of
  each extended partition in the chain is a partition
  table plus physically adjacent space, which is divided
  into the logical partitions as defined/referenced by
  that extended partition's partition table.

   the conflicts are whether or not you can (or should)
  have more than one logical per extended; and whether
  or not the pointer to the next extended in the chain
  must be the last entry in the (prior) partition table.

  QUESTION:  what constitutes “best practice‟ when using
           extended partitions?  what should I do, must
           I not do, and should be careful about doing?

   this posting is too long now, so I'll stop.
  many thanks for reading this far.  and thanks for any
  comments/answers/suggestions.

cheers!
	-blf-
--
«How many surrealists does it take to    |  Brian Foster      Montpellier,
 change a lightbulb?  Three.  One calms  |  blf at utvinternet.ie      France
 the warthog, and two fill the bathtub   |    Stop E$$o (ExxonMobile)!
 with brightly-colored machine tools.»   |        http://www.stopesso.com