Usefulness of SMART

[RichR]

Hello,

We have 135TB (45x3TB) in a 4U box running FreeNAS in a datacenter.

After reading this from some Google guys, then another thread that basically says SMART isn't reliable enough a mechanism to determine drive failure, I'm second guessing some things.....

An idea I had was to create/write a 1GB file, then checksum it, but all that would tell me is that the blocks I've written to are good/bad, the the overall health of the drive.

The questions are, do you solely depend on SMART to determine the health of your systems, and if not, what alternatives do you suggest.

NOTE: to clarify, 3 volumes, each with 14 drives + 1 warm spare in Z2

 

[louisk]

ZFS already does this internally. Just make sure you have the mail setup properly so it can notify you when there are pool issues.

 

[jgreco]

Probably best, at least if your data is even vaguely important to you, not to rely on the current mail setup, and make sure that you have your network monitoring system log in via ssh once an hour or so to do a zpool status and look for problems.

 

[Durkatlon]

I just wanted to say, that's a lot of space :D.

 

[louisk]

We have a similar setup at work, but it's 36x2T, configured at 4x 8x2T RAIDZ with 4 spares. I'm still working on testing performance, we just got some 10G deployed. Haven't had a chance to push the 10G yet, but the box will saturate 1G easily.