On Tue, 2005-06-07 at 16:46 -0700, Mike Hardy wrote:

Not sure about the size limits per se
(http://www.suse.de/~aj/linux_lfs.html has good info there)

Thanks for the link.

I can say that with a very large number of disks, you will hit
single-block read errors quite frequently due to the nature of the math
behind MTBF. This will lead to frequent drive expulsion / rebuild
cycles, as discussed in another thread today.

Say with 210 SATA's, for example?  Would I be replacing disks every day
or something?

In another question (you could post them all as one mail,
incidentally...)

True, but I've been hypothesizing that sometimes folks are reluctant to:

1) Wade all the way through a long series of questions

2) Respond to a message if they don't have the answers to all questions

 you asked if it was stable. The only evidence appears
to be anecdotal, but it appears extremely stable to me, even under very
heavy testing - normal new machine cpu/network/ram burn-in tests,
combined with multiple bonnie++ processes hitting the array. Any
problems I've had have been attributable to hardware or hardware drivers
thus far.

Good to know.

There have been problems (like the recent set of looping-resync issues
for instance), but I don't remember hearing any that resulted in data
loss, and they get fixed quickly. Serious problems still leave you with
your data on a set of disks in a known format, which means that it may
be laborious but you can reconstruct. So it "degrades" more gracefully
than the black box that are some hardware raid systems.

OK.

It takes care to run well and safely though, as with any machine.

Yes, I'm expecting that.

Thanks!