Best practice for filesystem layout in PXC


I have been playing with PXC and have had some issue with how I laid out my FileSystem. For starters I decided to create a separate mount point for /var/lib/mysql. This turned out to be a mistake as then I had an extra database for LOST+FOUND which caused all sorts of problems for my SST process. I tried to hide this database with the parameter ignore-db_dir=lost+found and while that worked it still caused a problem for the SST transfer where I would get a broken pipe. So by moving the datadir to another location all seemed to work.

So that lead me to lookup what might be the best practices be for a filesystem layout for a mysql database?

I came across the MOCA ( Mysql Optimal Configuration Architecture ) which seem reasonable but I am thinking that it might not be appropriate for a cluster db when dealing with backups and SST etc since it seems to spread things in many directories. So my question is how do you lay out a filesystem for PXC to avoid the most amount of problems.

I am running version 5.6 of PXC on debian 7.7


Some other topics I would like to talk about is best practices for installing the software? Does it make sense to install from the tar or use the .deb packages? I am thinking for upgrades it might be better to have multiple versions of the binaries in place and that way it is easy to revert if necessary.

What about using multiple NICs. I would have the eth0 for the public network where clients connect and eth1 for a private network where all cluster communication happens. Is this considered a best practice. I have seen some notes around but it seems like everyone uses different parameters to accomplish this. I would like a clear idea of what parameters to use.