Register for the iXsystems Community to get an ad-free experience and exclusive discounts in our eBay Store.

ECC vs non-ECC RAM and ZFS

Status
Not open for further replies.
Joined
Dec 5, 2011
Messages
5,952
Thanks
552
UFS will not prevent bitrot but it also will not cause further damage just from the RAM going bad and ZFS trying to "correct" the nonexistent errors.
ahh, didn't know that. if I didn't have ECC is still prefer zfs and just be happy im near sea level (less cosmic rays lol).
 
Joined
Mar 27, 2014
Messages
25
Thanks
1
I did memtest86+, no errors found.
So my setup is like this:

2 x RED WD HD 2TB: mirrored - UFS system
4xdiferent hds (6,5TB) in ZFS system.

The important data is on 2TB mirrored UFS system.

The ZFS volume is for media sharing only, so I can take the risk to loose that and I won't be so much worried.

I should let this nas box run like this until I buy some MB + MEM ECC.

Thanks
 
Joined
Mar 25, 2012
Messages
19,151
Thanks
1,857
ahh, didn't know that. if I didn't have ECC is still prefer zfs and just be happy im near sea level (less cosmic rays lol).
You realize that you can get radiation from the earth itself as well as the Sun. So being below sea level doesn't actually help you. ;) In fact, some of the most radioactive places on earth are at sealevel(beaches actually).
 

DJABE

FreeNAS Experienced
Joined
Jan 28, 2014
Messages
154
Thanks
1
Got my Crucial ECC RAM :) Still waiting for the MOBO thou.. xD
The odd thing with this pair of sticks is that the first one came out by China factory, and the second from the Puerto Rico factory. Even the PCB differs by color (China version is light green, while Puerto Rico is somewhat darker green...).
Even the manufacture dates are different. I thought I was buying paired sticks, it seems they just pick up on blind from their stock, and put it in a simple box with "2 pairs" printed on it..
 
Joined
Mar 25, 2012
Messages
19,151
Thanks
1,857
It won't improve anything. Some boards will run with ECC RAM, but they will operate as non-ECC RAM. But I definitely wouldn't buy RAM hoping it'll work. By a huge margin the number of motherboards that will use ECC RAM in a non-ECC condition is the minority.
 

HoneyBadger

Mushroom! Mushroom!
Joined
Feb 6, 2014
Messages
1,962
Thanks
652
It won't improve anything. Some boards will run with ECC RAM, but they will operate as non-ECC RAM. But I definitely wouldn't buy RAM hoping it'll work. By a huge margin the number of motherboards that will use ECC RAM in a non-ECC condition is the minority.
And they're pretty much all AMD boards, which are suggested against because by their nature they won't have an Intel NIC onboard.
 

DJABE

FreeNAS Experienced
Joined
Jan 28, 2014
Messages
154
Thanks
1
? AMD cpus does support ECC... but I'm not sure about motherboards thou... only ASUS claims their boards supports ECC too. But BIOS settings are very limited... you never really know does it work as full 72bit ECC or just it works as a regular 64bit system memory...
 

panz

FreeNAS Experienced
Joined
May 24, 2013
Messages
554
Thanks
43
On BSD Now!

http://www.bsdnow.tv

Allan Jude has just said he asked about ECC RAM to Matthew Ahrens and he told him... it's not so important :(
 

panz

FreeNAS Experienced
Joined
May 24, 2013
Messages
554
Thanks
43
Matthew Ahrens is (one) of the inventors of ZFS

http://mahrens.org/

I'm going to listen to that podcast again...
 
Joined
Mar 25, 2012
Messages
19,151
Thanks
1,857
I think we have to wait a day or two for it to be available. I can't find any way to scroll back and listen to the history.
 

D4nthr4x

FreeNAS Aware
Joined
Feb 28, 2014
Messages
95
Thanks
11
This video was recorded on April 1, he's clearly trolling.

edit: Just got to the part he is talking about. He doesn't grasp the complete picture and it isn't the guy that made ZFS in the video so we don't actually hear the words out of his mouth. tl;dr "Oh yeah at worst it will just rewrite the fixed data based on the checksum." What if the checksum is what gets corrupted by ram? etc. Clearly misinformed.
 

PenalunWil

FreeNAS Experienced
Joined
Dec 30, 2013
Messages
115
Thanks
12
I feel sorry for the half informed newbie that latches onto his words and buys a pile of non-ECC hardware... builds a ZFS based NAS and 12 months later... HELPPPP.

Is it possible to leave a comment at the bottom of the clip?
 

panz

FreeNAS Experienced
Joined
May 24, 2013
Messages
554
Thanks
43
Joined
Mar 25, 2012
Messages
19,151
Thanks
1,857
You are welcome to write him. Me personally I take this stance:

1. He didn't ask for my opinion. I don't think I need to tell him my opinion though since I have a nice detailed article on the topic.
2. I'm 99% sure he's read my article just because he made so many comments that pretty much came directly from my article.
3. I'm pretty sure he didn't mention me by name because he didn't want to give me any publicity(and I'm totally okay with that). I don't really want publicity or recognition for what I've done.
4. There is no "right and wrong" with this ECC vs non-ECC. There's "good" and "better". The question is where do you draw that line and what makes you think "good" is "good enough" for your situation. The whole reason I advocate for ECC RAM is because I don't think people realize how serious things get when you don't know how bad non-ECC RAM can be. Most people seem to have no experience with server-grade stuff. So helping them not make expensive money-sapping mistakes is a plus.
5. I'm most concerned with them knowing that this problem exists. If they still want to go with non-ECC RAM, that's fine. It's not my data and not my tears if something bad happens. I'm interested in them at least being able to make an informed opinion. If their informed opinion says non-ECC is fine, that's great. I'm glad they made that choice. I'd rather they make that choice knowing the potential risk than to not realize that non-ECC RAM can be bad. And let's face it.. 99% of us that show up in these forums don't know that non-ECC RAM can be so damaging.
6. Many people here will spend the extra $200 just to be 100% certain this doesn't happen to them. Most people that go with ZFS choose to go to it and accept the potential additional costs associated with getting compatible hardware. I fell into this category when I did my server. The cost was secondary to protecting my data, within reason.
 

panz

FreeNAS Experienced
Joined
May 24, 2013
Messages
554
Thanks
43
Joined
Mar 25, 2012
Messages
19,151
Thanks
1,857
Eh, that's totally their prerogative. Also Jude's answer clearly says "no more risk than UFS". He's absolutely right that there's "no more risk than UFS". The problem is that too many people here aren't wanting "the same risk as UFS". They want "lower risk than UFS". After all, if we all wanted "no more risk than UFS" we'd be using UFS, wouldn't we?
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,782
Thanks
3,036
That's a hard comparison to make; storing your files on a single disk with UFS is moderately risky. So I think I'll agree, literally, with that, but note that it is a substantial perversion of the issues to make non-ECC seem more acceptable. It will come as a shock to many people to discover that a ZFS pool that is merely being read can be corrupted by bad RAM, for example.
 

no_connection

FreeNAS Experienced
Joined
Dec 15, 2013
Messages
271
Thanks
6
So, does that mean they have fixed the risk of FreeNAS destroying the pool if a memory goes bad? Or if the system hangs, looses power or any other instability for that matter?
Does that mean that FreeNAS is now immune and that memory errors will not thrash the pool or risk data?

That would also mean that visualization is finally safe too with the new found stability.

Too bad I believed the risks where grounded in actual users loosing data when I bought my ECC based server. Could have saved me a few $ or cents if this news got to me earlier.

Sarcasm aside, it would be nice if data integrity and stability could be guaranteed without ECC, but most *possible* solutions points to some form error correction, which is what ECC does.
It would be odd to ask for exact quantities of water then give someone a bucket that may or may not have holes in it. Sure you know it's size and you can somewhat detect if it is leaking with enough buckets and a lot of work (or if the floor is getting wet) but it would be a lot simpler to have an integrity check verifying the content of the bucket every time you use it. That way you know it's right, or if it's wrong. And your feet would still be dry.
 
Joined
Mar 25, 2012
Messages
19,151
Thanks
1,857
This guy is an asshat.

edit: Don't trust anyone that isn't important enough to have a wikipedia page.
I don't have a wikipedia page.. thank God... I think.
 
Status
Not open for further replies.
Top