OK, maybe I have been making too much in the way of assumptions about the NetApp being any good (particularly at higher usage levels), but we got a clustered 7410 and it took over a year to accept it, and even then it was on the basis of us getting more stuff to make the acceptance tests go away.
I think ZFS is great, and in terms of I/O performance per pound, as well as GB/£, the Sun product was way better then the rest, but its utterly crap to administer!
Maybe you have had a good experience with a non-clustered version, but our problems started when it was delivered in April 2009 and would not boot, just hanging at the joining cluster stage! WTF was the testing like?
Since then we have had almost everything imaginable happen while testing and configuring the thing. Software updates that left it unbootable and forcing a version roll-back or factory reset. Features like saving the configuration that, when tested, trash the heads on restore (another factory reset), etc, etc.
Now we are using it for real, and we simply have to leave the settings in place for fear that something trivial, such as changing ownership of a network resource, will take the system off line for the best part of a day as we have to factory restore, then reconfigure everything, then hope it is working once more.
Oh yes, and we did lose data in one of our RAID rebuild tests, they say that was fixed in one of the 2.5.x firmware updates. This guy's summary kind of agrees with our experience, the sum being less than the parts:
We really really hope that Oracle will get it fixed properly, and soon, but fear the problems in the architecture of the akd stuff are too deep rooted to be easily fixed, and it will be dumped and a Mk2 brought out with a properly engineered version. Assuming they have not sacked all of their competent engineers by that time...