can RAM couse checksum error or is my drives faling?

[kaspernordmark]

Okay so here is the situation the system did a automatic scrub on my 2 mirrored 3tb drives a few days ago and ended up with maby 28 checksum on on of the drives and 7 on the other. And today I removed all of the errors with "zpool clear" and then i tryed to do a new scrub on the volume, wich just finnished and on of the drives say 16 checksum and the other say 43 checksum Repaired: 7.38M.

Do this indicate that my drives are about to fail or could it be the ram that I'm using?
I have used it (the ram) in my main pc before and whilst it was in my windows machine it crached at least 1 time every day and when I switched ram with the server (becouse the server neded 16gig and my win pc could make it with only 6gig) the windows pc stoped craching.

Sorry if I'm confusing but I'm verry new to FreeNAS and I'm swedish.

 

[zambanini]

did you read the faq? no, so a friendly reminder: read it, post infos regarding your hw und setup.

 

[kaspernordmark]

hw? My server setup is
Motherboard: ASUS M4A79T Deluxe, Socket-AM3
CPU: AMD Athlon II X2 245
Graphics: not sure but I think it is "ati 102-b17002(b)"
Ram: Corsair Vengeance DDR3 1600MHz 16GB CL9
Harddrives: 2 Seagate barracuda 3tb, 1 western digital 160GB
Power: enlight GPS-300AB-100 U 290Watt
OS runs from a 8GB Datatravel USB stick

 

[zambanini]

so no ecc ram. and a much to old cpu. when you really used the same ram modules before in the other maschine and it had random reboots and so on, it really sounds like a ram issue... you may get some sort of silent data corruption.
maybe @cyberjock will do a lol, just another case of unproper hw.

 

[kaspernordmark]

The thing is that all the components are "3rd generation" First my brother buys them brand new and uses them for a couple of years and then he buys new stuff and I get the old and when we repeat that process again the server gets an upgrade.

But on topic, I'm gonna try to change the RAM to Corsair Dominator DHX+ DDR3 1600MHz 6GB and see if it will make any difference.
I just wanted to know if RAM could couse checksum error.

 

[Ericloewe]

Non-ECC RAM is most certainly not ideal...

Now, post the output of smartctl -a /dev/adaWhatever for all drives (use pastebin please).

 

[cyberjock]

LOL!

 

[kaspernordmark]

Okay I changed the RAM and have been running it for a few days now without any error and I'm about to send the RAM back to the store. So I am really glad that I noticed this in time before I lost any major data and I have all to thank the freenas system with zfs.

I'm gonna use it like this for a month or so and if I don't get any more error I can officially say that foulty RAM can couse checksum error.

 

[cyberjock]

You don't need to officially say it. It's already known. If you've read that ECC vs non-ECC thread you'd know that you get checksum errors.

You'd also know that you also have potential corruption of files and the file system and it is impossible to detect it now.

You'd also know that what is likely to happen is things will be "okay" until one day the pool will crash and on reboot the pool never mounts again.

The best advice is to destroy the pool and restore from backup so you aren't hit with this ticking time bomb.

 

[9C1 Newbee]

If you've read that ECC vs non-ECC thread you'd know that you get checksum errors.

You'd also know that you also have potential corruption of files and the file system and it is impossible to detect it now.

You'd also know that what is likely to happen is things will be "okay" until one day the pool will crash and on reboot the pool never mounts again.

You should bind this text to your F5 key or 3rd mouse button.

 

[cyberjock]

Eh, not my problem. I said what I needed to say. The real question is whether the OP will realize he actually does need to destroy his pool. If he's like most others he'll:

1. Do a scrub of the pool.
2. See zero errors after the scrub finishes
3. Tell me I'm wrong and ask me to cite evidence (I won't do it because I have better things to do).
4. Will keep using the pool despite the warnings from me.
5. Some time will go by, then one day his shares (and his data) will suddenly disappear and he'll have evidence the pool is corrupted beyond recovery and won't mount.
6. He'll come back asking for help because it was the only copy of his data in existence.
7. We'll tell him "kiss the data goodbye". He'll try a bunch of stuff he found from Google searching, then eventually kiss the data goodbye.

 

[9C1 Newbee]

8. Claim FreeNAS is trash and say ZFS ate all his data.

 

[kaspernordmark]

I have almost all files copied but not all. Can I copy the rest of the files from the server to the backup? Or do i have to destroy them to?

Why wouldn't I listnen to you? I have followed your advice on a lot of other sstuff.

 

[kaspernordmark]

I would do a intel build witch ecc ram and all that if I only had the wallet to support it

 

[jkh]

Use the new backup / restore feature of FreeNAS 9.3. It's exactly this sort of Disaster Recovery scenario it was designed for.

 

[kaspernordmark]

So how do I use that?

 

[jgreco]

I would do a intel build witch ecc ram and all that if I only had the wallet to support it

So basically your options are:

1) You can continue spraying random bad bits into your pool
2) You can build a proper ECC build
3) You can fix the memory in your existing system and take your chances that it doesn't revert to situation 1).

You are paying a price NO MATTER WHAT YOU DO. If you can't afford to spend the money, it will cost you in bits lost (or at least bits risked). There is no fourth option.

 

[kaspernordmark]

I have already changed the ram in the system so how do I use the new backup/ restore feature? Or do what syberjock said?

 

[jgreco]

I don't know anything about the "new backup/restore feature" but the smart money's on getting the remaining data off your filer by copying it someplace safe. ZFS is going to be a really bad choice for you if your hardware is known to be faulty, because you'll never really know for sure if those "corrected" errors are actually being fixed properly or introducing new data errors into your data.

If your system is running smoothly with the new memory (and I strongly suggest your run MemTest86 on it for a week or two before even thinking of making such a claim), you have to make a call as to whether or not you want to trust it moving forward. If so, you need to destroy and recreate your pool, because if corruption was introduced into your pool by bad RAM, you have a potentially ticking time bomb... even if the memory is better *NOW*, corruption that happened last week doesn't magically get fixed by the new memory, and the pool is still damaged.

 

[kaspernordmark]

Yes I think I'm gonna do what you said but can I/ is it safe to copy all the movies that I have on the server to a backup before destroying the pool? And do I have to reinstall freenas on the boot drive? (USB stick)