Question for the hardware gurus- about NON EEC ( duh duh duh........ )

[Mega Man]

please note. this has NOTHING to do with my production machine.... except i am a) out of sata ports atm and b) need to test new hdd, and when test/burn in is complete i will remove 1 disk from my current pool and replacing it with a new disk, for an entirely new pool and basically coping my old pool to my new pool ( with additional disk - going from raid z3:6 disk to raid z3:7disk pool as per cyberjock's thread 7 disk is superior then 6 disk for z3 ( or i may buy a lsi 9211-8i, because i can and use that !..... after burn in ))

that said my current freenas build is 100% irrelevant to my question, hence the lack of equipment list.

if i build a freenas box, from my leftover parts, 100% consumer grade non eec, for just doing HDD burn in/testing is there anything that can fail due to that ??? ( IE the tests would they be invalid ? )

i dont see how as i am testing the writing of the hdds and comparing the results to a known list, but i just am paranoid and want to make sure i am 100% ok.

just to reiterate this the NON eec test build will NEVER be used in production --- and/or can i use this thread / resource and will it work on my PF sense box, which also has 32gb EEC !--( tested- and nope.... )

THANKS SO MUCH !!! for your time !~

 

[jro]

Non-ECC would be fine for a testing/burn-in machine. In general, ZFS running on a system without ECC memory is no more dangerous than most other file systems running on ECC memory. This is thanks to ZFS's end-to-end data verification capabilities.

By the way, a 7-wide Z3 vdev is not necessarily "superior" to a 6-wide Z3 vdev (other than the fact that it has more spindles, so throughput will be higher). The whole "data disk count should be a power of 2" rule has kind of fallen by the wayside thanks to the ability to tune your recordsize value. More info here: http://jro.io/nas/#overhead

 

[Mega Man]

thanks!

as to the number of drives, i cant really comment as i dont quite understand why, i can state i saw it here on cyberjocks Slideshow explaining VDev, zpool, ZIL and L2ARC for noobs!

on page 34, again it says it can be disregarded, but meh why not ...., as to the recordsize value, thanks ill have to look into that !

 

[danb35]

Back in the day, there was a guideline that the number of disks in RAIDZp should be 2^n+p--some power of two plus the number of parity disks, which would recommend that a RAIDZ3 would be better as 7 disks than as 6. With the use of ZFS compression, that guideline has pretty well gone out the window.

 

[Mega Man]

Thanks! That helps allot

 

[Ericloewe]

Matt Ahrens has a spreadsheet that actually tells you how much padding each RAIDZ width needs for typical block sizes. It's not very important at all if you're using large files.