[lug] Reliable SATA server?

[Rob Nagler]

Hi Sean,

> We've had probably 40 of these through either our facility or our clients,
> and have been very happy with them.

I guess it would be good to know the exact configuration, including
which chassis.  Heat is a big problem in computers nowadays.  The only
reason I don't like Dell is that I can't get enough disk slots in
their lower end boxes.  It costs a lot to get 12 drives chassis.

> I think you'd be hard pressed to call using filesystem snapshots "more
> expensive and complex", but whatever.  :-)

ZFS is more complex than ext3 by far.  While it may be engineered very
well, it still has more moving parts than ext4, and many more moving
parts than cp.  ZFS is also very new.  There have been very serious
bugs in it quite recently.  It doesn't come stock on CentOS, and I'm
using multiple VM vendors plus my own machines.

Yes, I'm very conservative.  I'm also lazy.  I program in Perl after
all. ;-)   I'm not a big fan of the latest and greatest technology.  I
am running Lion for me, and it has had a horrible reliability record
for me.  I just bought a Seagate SDD Hybrid, and it failed within
weeks.  I should have bought a plain old 750GB drive, instead of
trying to get fancy, thinking it would be "faster", when in fact, it
has cost me hours in lost work, which could never be made up by a few
milliseconds per seek that the drive would net me.

> In our environment "cp -l" wouldn't work at all.  It uses too much space

Yes, to each his own.... My solution scales, because I have control
end-to-end.  You don't have the control I do over the applications.
I'm not running systems for other people so I get to pick and choose
what data I back up, and nobody runs MH or netnews on our servers. ;-)

> another system.  So you can do the replication to a secondary backup server
> without having to compare the whole file-system on both sides.

This is interesting for HA.  Mixing threads.  How fast can ZFS compare
a tree with 1M+ files in about 100GB?

> Sorry I couldn't be of help.

Oh, you've given me ideas. :)

> With all due respect, the above statement to me sounds like you are
> unwilling to even consider that "cp -l" is not a good way to store historic
> backup copies.

"good" is a funny word, isn't it?  In my value system, "cp --link" is
a great way store historic copies.  In your value system, it's a waste
of resources, and wouldn't work for your environment.  That certainly
qualifies as "bad" to me.

> And if you aren't willing to consider that, there's not
> much any of us can do.  You've apparently been working with computers for
> too long to accept my 25 years of experience...

I didn't mean to say that you didn't know what you are talking about.
To the contrary, I think you have great experience (as do others on
this list).  We both have valid experience, no matter how many years,
and we certainly have translated our experience into successful
businesses.  My point is that saying "it's the user's fault" is just
plain wrong.  I educate our users all the time to appreciate that the
system should do what they expect, not what the programmer expects.

Rob