Re: WAPBL and write cacheing

[Mouse <mouse%Rodents-Montreal.ORG@localhost>]

>> I learned that it's safe to use write cacheing on a SATA drive with
>> FFS+WAPBL.

>> I would like to know
>> 1. whether it's safe on SAS drives, too
>> 2. whether it's safe with plain FFS, too

That depends heavily on what you consider "safe".  Your data will never
be absolutely safe; you just have to decide where the "beyond this the
costs outweigh the benefits" line is for you.  Basically, there is no
"safe", only "safe enough".

For example, I use plain FFS on a wide variety of drives - IDE, SATA,
SCSI, from mutliple manufacturers and multiple generations of
technology - and don't worry about it, largely because I have
sufficient backup mechanisms in place that a drive dying entirely is an
annoying but hardly catastrophic event.  (I know this because I've lost
some four or five drives since I got my current paradigm set up; all
the bits were intact each time.)  I've seen numerous ungraceful
powerdowns on FFS filesystems in my 25+ years as a sysadmin, and I can
count on one hand the number of filesystems corrupted beyond the
ability of fsck -p to recover by such an event.  (I've seen plenty of
corruption, but in almost all cases I've identified other causes for
it, from overlapping partitions to hardware failures.)

So, for my purposes - personal machines, downtime is annoying but
nothing worse, and there is always the skill and knowledge available to
restore from backups and/or repair filesystem damage - plain FFS with
no other precautions beyond good backups is "safe enough".

Only you can decide how far you'll have to go to be safe enough for you.

> I would say plain FFS is only ok with write cacheing if you have
> stable power.  Which means not really, but everybody does it anyway :-)

Well...I'm not sure I'd say "not really".  FFS is remarkably robust in
practice even in the absence of theoretical guarantees - guarantees
which depend on the absence of things like hardware failures anyway -
so, in many cases, even things like WAPBL are overkill in the sense
that the chance of a data-losing failure is already small enough.

/~\ The ASCII                             Mouse
\ / Ribbon Campaign
 X  Against HTML                mouse%rodents-montreal.org@localhost
/ \ Email!           7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B