On Sun, May 3, 2020 at 10:56 AM Caveman Al Toraboran <
toraboracave...@protonmail.com> wrote:
>
> On Sunday, May 3, 2020 1:23 PM, Wols Lists <antli...@youngman.org.uk>
wrote:
>
> > For anything above raid 1, MAKE SURE your drives support SCT/ERC. For
> > example, Seagate Barracudas are very popular desktop drives, but I guess
> > maybe HALF of the emails asking for help recovering an array on the raid
> > list involve them dying ...
> >
> > (I've got two :-( but my new system - when I get it running - has
> > ironwolves instead.)
>
> that's very scary.
>
> just to double check:  are those help emails about
> linux's software RAID?  or is it about hardware
> RAIDs?
>
> the reason i ask about software vs. hardware, is
> because of this wiki article [1] which seems to
> suggest that mdadm handles error recovery by
> waiting for up to 30 seconds (set in
> /sys/block/sd*/device/timeout) after which the
> device is reset.
>
> am i missing something?  to me it seems that [1]
> seems to suggest that linux software raid has a
> reliable way to handle the issue?  since i guess
> all disks support resetting well?
>
> [1] https://en.wikipedia.org/wiki/Error_recovery_control#Software_RAID
>

When doing Linux RAID, hardware or software, make sure you get a RAID aware
drive that supports TLER (Time Limited Error Recovery) or whatever the
vendor that makes your drive calls it. Typically this is set at about 7
seconds guaranteeing that no mater what's going on the drive will respond
to the upper layers (mdadm) to let it know it's alive. A non-RAID drive
with no TLER feature will respond when it's ready and typically if that's
longer than 30 seconds then the RAID subsystem kicks the drive and you have
to re-add it. While there's nothing 'technically' wrong with the storage
when the RAID rebuilds you eventually hit another on of these >30 second
waits and another drive gets kicked and you're dead.

I've used the WD Reds and WD Golds (no not sold) and never had any problem.

Build a RAID with a WD Green and you're in for trouble. ;-)))

HTH,
Mark

Reply via email to