ECC Memory Question

[Kelly Shutt]

First, please don't bring up the ECC vs. non-ECC argument here. I do understand the risks of running non-ECC memory. In fact, that is part of the reason I'm making this post.

I am currently in the situation where I have a running FreeNAS system with non-ECC memory. I am in the position to be upgrading this system soon and I'm considering switching over to ECC memory in the upgrade. I have the money and the system has grown to the point that I no longer consider this as acceptable risk.

Knock on wood here... This system has been running for several years now, and I haven't experienced any evident corruption due to memory issues. This system is kept up to date with the latest stable FreeNAS release; including keeping the zpools up to date with current FreeNAS standards.

The question is this... Are there considerations I should make when upgrading the system with ECC memory. Is it possible that by installing ECC memory it will suddenly start detecting corruptions that were not detected previously? Or, at this point should I consider things to be (mostly) intact and it would be safe to make the change? My zpools are large enough that it would be quite difficult for me to back them up before the upgrade, so I'm hoping I can just put in the new hardware and go.

Thanks,
Kelly Shutt

 

[Kelly Shutt]

If anyone is interested, I have a list of the CPU, Memory, and Motherboard that I'm considering.

http://secure.newegg.com/WishList/PublicWishDetail.aspx?WishListNumber=23201311



Peace,
Kelly Shutt

 

[BigDave]

My zpools are large enough that it would be quite difficult for me to back them up before the upgrade

Would this be more difficult than attempting to recover your data after your pool fails during the upgrade?
As a Network Systems Admin, I would think you should know better. To put it bluntly, you're playing
Roulette with all chambers loaded at this point. If you value your data, backup your pool before attempting
this so we don't have to read about another pool going down the drain. Good luck.

 

[Ericloewe]

Moving from non-ECC to ECC doesn't make errors appear.

Chances are, if everything's working fine, your pool has survived this far. It's not guaranteed, since some crazy coincidence could easily mask any corruption that might have happened.

If you're staying on the same version, move over the pool and the boot device and it should be plug an play, except maybe some networking stuff.

 

[SmallGuy]

I have play that game last week.
I have been lucky, the gun was stuck! ;)
In fact all my valuable data were back-up. Only my movie "collection" wasn't.
But the hardware upgrade has been flawlessly done.
Good luck!

 

[joeschmuck]

The beauty of FreeNAS is it's hardware independence. You can move your drives over and your boot USB Flash to other reasonable hardware and it should work without issue. As for if you add ECC RAM and suddenly you may find data errors, that is not the way ECC RAM works and you apparently have not done much research on what ECC RAM does nor why ECC RAM is important in a ZFS system. As a System Administrator I would think that you would want to know the details on how your system works but maybe you are part of a younger generation who doesn't care about how it works, just so long as it works enough.

 

[SmallGuy]

May be the ECC RAM will throw one disk outside the case. I hope your pool is RAIDZ2, and you will be able to survive. :P

 

[qwertymodo]

Any errors caused by bit flips in non-ecc ram are completely undetectable after the fact, that's the whole reason they're so insidious. For that reason, your system won't magically "notice" old errors when you switch to ecc.

 

[jgreco]

Any errors caused by bit flips in non-ecc ram are completely undetectable after the fact, that's the whole reason they're so insidious. For that reason, your system won't magically "notice" old errors when you switch to ecc.

That's basically correct. If you were monitoring the pool status all along and never saw checksum errors, that's about as good as it gets - it is no guarantee that there is no corruption, but it places a limit on the types of corruption that could have happened.

Of course, most people don't monitor their pools closely enough to notice anyways.