ZFS in production enviroments

[Mark Felder]

On Thu, 04 Apr 2013 08:02:04 -0500, dennis berger <db at nipsi.de> wrote:

> I case of a failure do you switch the head units manually?!

This was not designed to be HA storage. If we wanted to do that I'd have  
figured out how to bring HAST into the mix but HAST+ZFS is very messy.  
However, if a head unit completely died we'd just connect the JBODs to the  
other head unit and import the pool (you can daisy-chain them if you're  
out of ports on the controller). However, we do as a general rule stay  
under 50% storage so if we really needed to we could move all data to the  
other server without customers knowing, perform maintenance, and move data  
back.