[wellylug] Testing hard disks

[Ewen McNeill]

In message <20030917083749.DB8563C486E4 at basilica.la.naos.co.nz>, Ewen McNeill writes:
>In message <20030917054042.GA30107 at paradise.net.nz>, Volker Kuhlmann writes:
>>A 15G disk I have is throwing I/O errors on the 5G partition, all the
>>rest is fine. I nulled out that partition with dd and then ran badblocks
>>-w, which found no errors. When I went to install the system on that
>>partition, the I/O errors in syslog keep scrolling past.

One more thought: if the errors appear when reading from the disk but
disappear once you've written to the disk and read back again, then I
suspect the disk is actually doing "sector relocation" on you -- ie,
instead of using the sectors it has found "bad", it's using another set
of sectors in its relocation area for them.

Since that pattern has been happening repeatedly for you with that disk
(ie, can't read from areas of it, write and it's okay for a while,
then can't read/write to areas of the disk), I strongly suspect that
there's a significant and growing area of the disk which is no longer
usable and if the relocation area isn't already used up it soon will be.

Thus it might be better to avoid that disk.

SCSI disks generally provide a way to retrieve the list of relocated
sectors (and often to update it); IDE disks generally don't.  IDE disks
are generally just "magic".

Even if that isn't the case, if the disk keeps "going bad" after
attempts at blanking it out, newfsing it, etc, then I'd be dubious about
using it.  Especially with a 15GB disk which is relatively cheaply
replaced these days.

Ewen