This is the FreeNAS forum. I'm not looking to provide a dual head solution.
If I were looking to provide a dual head solution, I'd *still* note that the ZeusRAM is three year old, pricey, 3.5", SAS based technology by a manufacturer that has been bought out, first by HGST, then by WD, so in all likelihood it isn't going anywhere.
But we knew that this discussion wasn't about a dual head solution from the start, since we were discussing abusing a RAID controller's cache as an alternative low latency SLOG, and that precludes any reasonable dual head solution.
ZFS isn't real strong in the dual head department anyways. Nexenta's strategy is to monitor availability and forcibly import a pool and restart services on the backup head, a process which makes me vaguely uncomfortable because we know forcible pool imports are a dodgy thing, and the risk of damage from a split brain scenario is pretty horrifying. I know Nexenta has rudimentary guards against that, but when it comes right down to it, the possibility still exists. A lot of what they do is somewhat obfuscated at the marketing and sales engineering levels, and there are some glaring technical deficiencies when you take a harder look, such as the
failure to detect NFS/CIFS service failures (page 16). A lot of the hype started seeming a lot more believable once it was recognized as "feature checklist" sales hype rather than hard engineering achievement.
So even with SAS dualporting, your pool is still a SPOF ("power event!"), and right now I'm not aware of a good way to accomplish SPOF-free ZFS.
While you're fine with a SPOF, I'm not.
I appreciate how you might have come to your errant conclusion about what I'm fine with. I tend to build redundancy at higher levels, assuming catastrophe will strike random lower levels. I'm an
early advocate of RAIS.