«608 »
  • Post
  • Reply
D. Ebdrup
Mar 13, 2009



I have linked it several times, here is the explanation for why you should use ECC for zfs.

D. Ebdrup fucked around with this message at 18:01 on Mar 22, 2014

evol262
Nov 30, 2010
#!/usr/bin/perl


No offense, but both this and the previous are "why you shouldn't trust software RAID". The entire point of ZFS' self-healing is that it performs better in loss situations than other software RAID, and as close to hardware parity calculations and battery-backed caches as you can get to on plain hardware. Obviously ECC is better, but not having ECC does not make it worse than mdraid, storage spaces, or dmraid, matrix raid, or any other software system. It's still better. You don't need ECC. Stop sperging about it.

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

evol262 posted:

No offense, but both this and the previous are "why you shouldn't trust software RAID"...You don't need ECC. Stop sperging about it.
If all you're doing is storing movies and easily replaceable shit on your NAS, then you are correct--if you memtest everything before filling it with data, the chance of running into bad RAM is reasonably low. However, in that a basic ECC setup doesn't add too much to the cost ($125 for a motherboard, $100/8GB ECC RAM versus $100 for a motherboard and $75/8GB for non-ECC RAM--so you're looking at ~$55 more for an average setup, and $25/8GB RAM more past that), it makes sense to stress it for everyone who either has difficult to replace data (~*~RAID Is Not A Backup~*~), or who values their time more than their $55.

evol262
Nov 30, 2010
#!/usr/bin/perl

DrDork posted:

If all you're doing is storing movies and easily replaceable shit on your NAS, then you are correct--if you memtest everything before filling it with data, the chance of running into bad RAM is reasonably low. However, in that a basic ECC setup doesn't add too much to the cost ($125 for a motherboard, $100/8GB ECC RAM versus $100 for a motherboard and $75/8GB for non-ECC RAM--so you're looking at ~$55 more for an average setup, and $25/8GB RAM more past that), it makes sense to stress it for everyone who either has difficult to replace data (~*~RAID Is Not A Backup~*~), or who values their time more than their $55.

Which is what 99% of people do. But that aside, ZFS isn't unreliable without ECC. And ECC doesn't save you from disk errors, which are far more likely to eat your data. If your $55 makes you sleep better, great, but literally every operation taken (including data transfer) would be affected by the "stuck bit" problem, and they'd core dump, and you'd notice. It's not silent. ECC buys very little for a home NAS. Again, if it gives you peace of mind, OK. But presenting it as disastrous or irresponsible to not use ECC is scaremongering.

adorai
Nov 2, 2002

10/27/04 Never forget

Grimey Drawer

if you are storing actual irreplaceable data that cannot have errors (i.e. financial transactions) you would be dumb to not use ECC. If on the other hand you are filling it with movies, music, photos of cousin jimmy and a few word docs, ECC is probably overkill. The ram itself isn't that much more expensive, but the motherboard will be.

fletcher
Jun 27, 2003

ken park is my favorite movie

Cybernetic Crumb

Trying to clone a dying hard drive to the replacement they sent me. Everything on the drive is already backed up, I just wanted to save myself the hassle of reinstalling a bunch of crap. I selected "interactively check and repair source file system before cloning" but that doesn't seem to be the same thing as the --rescue option mentioned in the screenshot? Not sure how to proceed here.

IOwnCalculus
Apr 2, 2003





The problem is that your drive is failing in such a way that it's puking errors and failing to respond to commands. A lot of times when drives do that, they go unresponsive for a long time.

I'd chalk up any attempt to recover data from that drive as a loss and move on, you'll spend less time restoring from backup than trying to wrestle that.

fletcher
Jun 27, 2003

ken park is my favorite movie

Cybernetic Crumb

IOwnCalculus posted:

The problem is that your drive is failing in such a way that it's puking errors and failing to respond to commands. A lot of times when drives do that, they go unresponsive for a long time.

I'd chalk up any attempt to recover data from that drive as a loss and move on, you'll spend less time restoring from backup than trying to wrestle that.

Sounds good, thanks for the advice!

D. Ebdrup
Mar 13, 2009



If you've got time and freebsd (or one of it's live-cd variants, or another OS that it has it as it in binary form), throw recoverdisk at it. No guarentees, but it was designed by Poul-Henning Kamp (one of the core FreeBSD developers) designed to recover the maximum amount possible (eg. everything remotely recoverable) by reading everything it can and marking which blocks can't be read, then going back and trying to read those previously unreadable blocks again and marking the unreadable ones, then rince and repeat.

EDIT: ↓ Carbon Copy Cloner has a freeware version that'll let you copy data from one location to another, if that's what you're looking for. However, HFS+ - while supporting journaling - doesn't support per-block checksumming, and I don't quite see how IntegrityChecker is supposed to know the original checksum of every block or file on the disk.

D. Ebdrup fucked around with this message at 21:35 on Mar 23, 2014

the_lion
Jun 8, 2010

On the hunt for prey...

I'd love to use zfs, but it's a bit out of my reach-mac drivers are buggy apparently.

Bit of a cross post (asked in mac thread) but is there a good tool for checking OS X file integrity when backing up?

Google leads me to this:
http://diglloydtools.com/integritychecker.html

skooma512
Feb 8, 2012

You couldn't grok my race car, but you dug the roadside blur.


WD Caviar Blue is officially off my buy list.

I have 3 of those suckers on my desk at work. All are bricked because whole drive encryption was too much for them. They are all only moderately used and less than 3 years old.

withak
Jan 15, 2003


Fun Shoe

Due to working too much, I have a "bonus" from my employer where I can get something from amazon for up to $250 (including California sales tax, so up to about $230 really). I'm looking at getting either a Synology DS213j or a QNAP TS-220 which both seem to do approximately the same things and fit in the price range.

Is there any consensus on which is better? Is there anything else that I should be considering?

http://www.amazon.com/QNAP-TS-220-P...4&keywords=qnap

http://www.amazon.com/Synology-Disk...95764540&sr=1-7

Thanks Ants
May 21, 2004

Bless You Ants, Blants



Fun Shoe

They are both fairly low powered single-core boxes, but will be more than fine as a NAS as long as you aren't planning on doing a load of video transcoding on them. If you are planning to use it for something more advanced than plain NAS then the prices for beefier hardware get very high very quickly so you're almost better off rolling your own Xpenology box at that point.

Qnap vs. Synology is personal preference a lot of the time - I've only ever used Synology and have been very happy with the product and their tech support, so haven't had a reason to look elsewhere.

withak
Jan 15, 2003


Fun Shoe

Yeah I figured that vanilla NAS stuff is all that would fit in the $230 limit. I already have Plex running on my computer, it would continue to work if I moved all of my library stuff from my HD over to one of these things, right? Is there much of a performance hit if Plex on a computer is pulling its source files from one of these instead of from the local HD?

withak fucked around with this message at 00:30 on Mar 26, 2014

LRADIKAL
Jun 10, 2001

A Very Useful Person



Fun Shoe

skooma512 posted:

WD Caviar Blue is officially off my buy list.

I have 3 of those suckers on my desk at work. All are bricked because whole drive encryption was too much for them. They are all only moderately used and less than 3 years old.

Seems more likely to me that you got part of a poorly shipped or part of a bad run. They're still under warranty, right?
Has anyone else had issue with Blues? I have one with 4.68 years of uptime!

DrDork
Dec 29, 2003
commanding officer of the Army of Dorkness

Jago posted:

Seems more likely to me that you got part of a poorly shipped or part of a bad run. They're still under warranty, right?
Has anyone else had issue with Blues? I have one with 4.68 years of uptime!
Blues are still decent general-purpose drives. As you say, no matter what drive line you pick, someone is going to pipe up with unfortunate experiences. The 2 year warranty on Blues, though, mean he's probably out of luck on that front.

skooma512
Feb 8, 2012

You couldn't grok my race car, but you dug the roadside blur.


Jago posted:

Seems more likely to me that you got part of a poorly shipped or part of a bad run. They're still under warranty, right?
Has anyone else had issue with Blues? I have one with 4.68 years of uptime!

They're through Dell so I don't know if they are under the computer's warranty or a separate one. I hesitate to RMA them because I fear the inevitable cries of MY DATA (this is why we gave you server storage).

Still pretty damning that they are the only ones that have problems. I found even more in our storage room. Anything 250GB and above is all Blues.

To think they were on my Amazon list. I was going to replace my aging but still working HDs with them

Tsuru
May 12, 2008


evol262 posted:

Which is what 99% of people do. But that aside, ZFS isn't unreliable without ECC. And ECC doesn't save you from disk errors, which are far more likely to eat your data. If your $55 makes you sleep better, great, but literally every operation taken (including data transfer) would be affected by the "stuck bit" problem, and they'd core dump, and you'd notice. It's not silent. ECC buys very little for a home NAS. Again, if it gives you peace of mind, OK. But presenting it as disastrous or irresponsible to not use ECC is scaremongering.
I'd be a lot more worried about cheap SATA adapters or cables fucking up my data than lack of ECC, anyway. Out of all the adapters I have bought over the last couple of months, I have 1 (one) which works without puking cksum errors. The rest, including 2 SiI3114 based cards and one LSI which worked really well for about 10 minutes before it fried itself have been nothing but a pain in the ass.

To be clear: that's a 80% failure rate. Serves me right for ordering only the cheapest crap off of ebay, I guess.

MMD3
May 16, 2006

Montmartre -> Portland

I got my synology 1513+ in last night and installed the 3 3TB Red drives I got for it.

7 hours later and the parity consistency check is only at 70% complete. *sigh*

guess that's just how things are going to go when you're working with 9TB of storage.

ShaneB
Oct 22, 2002



MMD3 posted:

I got my synology 1513+ in last night and installed the 3 3TB Red drives I got for it.

7 hours later and the parity consistency check is only at 70% complete. *sigh*

guess that's just how things are going to go when you're working with 9TB of storage.

I think it took over 24 hours to add a 3rd 3TB drive to my array.

codo27
Apr 21, 2008

"I dont fully understand football contracts but you can just be outright cut if you're shit right? With no penalties? Hockey needs that."

I Am Marc Bergevin IRL


I see the QNAP mentioned in the OP, and that's what we use here at work (government). But TigerDirect has the Buffalo 421e on for $109 CDN, tax and shipping all told I'm looking at $140 and change, and just after a few quick quotes that's about $20 cheaper than I can get the QNAP 212-e. Should I spend the extra cash?

MMD3
May 16, 2006

Montmartre -> Portland

ShaneB posted:

I think it took over 24 hours to add a 3rd 3TB drive to my array.

Oh jeez, okay, had no idea it took that long. No wonder people do SHR2

ShaneB
Oct 22, 2002



MMD3 posted:

Oh jeez, okay, had no idea it took that long. No wonder people do SHR2

Hold on, what are you using?

MMD3
May 16, 2006

Montmartre -> Portland

SHR1?

Ninja Rope
Oct 22, 2005

Wee.


Anyone else get "Time Machine completed a verification of your backups. To improve reliability, Time Machine must create a new backup for you" backing up via Time Machine to ZFS via Netatalk? fscking the actual sparsebundle shows real errors, but I don't know how the corruption occurs in the first place.

I'm not using FreeNAS, but the FreeNAS docs suggest it's caused by doing a zfs scrub but I don't understand why a scrub would corrupt data like that. All of my scrubs report 0 errors found/repaired.

Any suggestions on what I can do to fix this? I'm using netatalk 2.2.5 and I could try upgrading to 3.1 but nothing in the changelog seems to relate to this.

Crossbar
Jun 16, 2002
Chronic Lurker

Ninja Rope posted:

Anyone else get "Time Machine completed a verification of your backups. To improve reliability, Time Machine must create a new backup for you" backing up via Time Machine to ZFS via Netatalk? fscking the actual sparsebundle shows real errors, but I don't know how the corruption occurs in the first place.

I'm not using FreeNAS, but the FreeNAS docs suggest it's caused by doing a zfs scrub but I don't understand why a scrub would corrupt data like that. All of my scrubs report 0 errors found/repaired.

Any suggestions on what I can do to fix this? I'm using netatalk 2.2.5 and I could try upgrading to 3.1 but nothing in the changelog seems to relate to this.

I've had these errors and never figured it out. Please post if you do.

dox
Mar 4, 2006


MMD3 posted:

Oh jeez, okay, had no idea it took that long. No wonder people do SHR2

Please, for the love of all things almighty, use SHR2. This is coming from someone with an 1812+ 12TB array (2x 4TB, rest 2TB) who is stupidly using SHR1. I cross my fingers the entire ~24 hours of rebuild time when I add a new drive, praying to anyone who will listen that another drive doesn't fail during that rebuild...

And on a side note, rebuild times will still be the same/similar on SHR2... unless I'm missing something obvious.

ShaneB
Oct 22, 2002



dox posted:

Please, for the love of all things almighty, use SHR2. This is coming from someone with an 1812+ 12TB array (2x 4TB, rest 2TB) who is stupidly using SHR1. I cross my fingers the entire ~24 hours of rebuild time when I add a new drive, praying to anyone who will listen that another drive doesn't fail during that rebuild...

And on a side note, rebuild times will still be the same/similar on SHR2... unless I'm missing something obvious.

You can't SHR-2 3 drives. You would need 4.

Legdiian
Jul 14, 2004


dox posted:

Please, for the love of all things almighty, use SHR2. This is coming from someone with an 1812+ 12TB array (2x 4TB, rest 2TB) who is stupidly using SHR1. I cross my fingers the entire ~24 hours of rebuild time when I add a new drive, praying to anyone who will listen that another drive doesn't fail during that rebuild...

And on a side note, rebuild times will still be the same/similar on SHR2... unless I'm missing something obvious.

I just put together a 1513+ with (5) 4TB drives. I believe I used SHR1. Is it possible to switch to SHR2 without moving the data off and back on? Would I lose any usable storage?

Don Lapre
Mar 28, 2001

If you're having problems you're either holding the phone wrong or you have tiny girl hands.


Legdiian posted:

I just put together a 1513+ with (5) 4TB drives. I believe I used SHR1. Is it possible to switch to SHR2 without moving the data off and back on? Would I lose any usable storage?

You will have to move your data off, resetup as SHR2 and move your data back. You will lose one drive worth of space.

MMD3
May 16, 2006

Montmartre -> Portland

ShaneB posted:

You can't SHR-2 3 drives. You would need 4.

I hate you ShaneB... are you saying i have to go order another 3TB drive to sacrifice to the file safety god? HOW ABOUT YOU BUY ME ANOTHER DRIVE

eddiewalker
Apr 28, 2004


MMD3 posted:

I hate you ShaneB... are you saying i have to go order another 3TB drive to sacrifice to the file safety god? HOW ABOUT YOU BUY ME ANOTHER DRIVE

Or spend less than the cost of a drive for an off-site backup of your actual important files. In the unlikely event of a second drive failing during rebuild, everything else is replaceable.

MMD3
May 16, 2006

Montmartre -> Portland

eddiewalker posted:

Or spend less than the cost of a drive for an off-site backup of your actual important files. In the unlikely event of a second drive failing during rebuild, everything else is replaceable.

well... I'll be doing that for my raw photo assets anyways... which is why I was less concerned.

I think I'm just going to go ahead and keep SHR-1

spoon daddy
Aug 11, 2004
Who's your daddy?

College Slice

Ninja Rope posted:

Anyone else get "Time Machine completed a verification of your backups. To improve reliability, Time Machine must create a new backup for you" backing up via Time Machine to ZFS via Netatalk? fscking the actual sparsebundle shows real errors, but I don't know how the corruption occurs in the first place.

I'm not using FreeNAS, but the FreeNAS docs suggest it's caused by doing a zfs scrub but I don't understand why a scrub would corrupt data like that. All of my scrubs report 0 errors found/repaired.

Any suggestions on what I can do to fix this? I'm using netatalk 2.2.5 and I could try upgrading to 3.1 but nothing in the changelog seems to relate to this.

If the spare bundle is screwed, you can try this.

http://www.garth.org/archives/2011,...kup-errors.html

ShaneB
Oct 22, 2002



MMD3 posted:

I hate you ShaneB... are you saying i have to go order another 3TB drive to sacrifice to the file safety god? HOW ABOUT YOU BUY ME ANOTHER DRIVE

Hey man I run 3x3TB I assume I won't lose 2 drives at once. I also remote backup my important files from the NAS as well. If I lose my digital TV and Movies, life goes on somehow. If I lose my digital music and photos going back to 2005... stuff is a lot sadder.

Ninja Rope
Oct 22, 2005

Wee.


spoon daddy posted:

If the spare bundle is screwed, you can try this.

http://www.garth.org/archives/2011,...kup-errors.html

Thanks, I've seen this around but unfortunately this won't fix the underlying cause. In my case I get corruption every few weeks and I'd rather track down the cause than repair the bundle every time.

IOwnCalculus
Apr 2, 2003





ShaneB posted:

Hey man I run 3x3TB I assume I won't lose 2 drives at once.

This is a bad assumption. When rebuilding, you don't need a full-on shit from one of the other two drives - just an unrecoverable read error, and that's more likely than you want to admit when reading a whole 3TB drive.

eddiewalker
Apr 28, 2004


IOwnCalculus posted:

This is a bad assumption. When rebuilding, you don't need a full-on shit from one of the other two drives - just an unrecoverable read error, and that's more likely than you want to admit when reading a whole 3TB drive.

If a URE is encountered during the rebuild, I hope that doesn't stop the whole process, right? A corrupt spot in an audio or video file causing a skip, or a black line in a photo is something I can live with.

Kreeblah
May 17, 2004

INSERT QUACK TO CONTINUE



Taco Defender

I had an SSD fail in my ZFS pool and I'd like to replace it, but, uh:

pre:
  pool: zpool0
 state: ONLINE
status: One or more devices has experienced an unrecoverable error.  An
	attempt was made to correct the error.  Applications are unaffected.
action: Determine if the device needs to be replaced, and clear the errors
	using 'zpool clear' or replace the device with 'zpool replace'.
   see: http://illumos.org/msg/ZFS-8000-9P
  scan: scrub repaired 0 in 4h15m with 0 errors on Thu Feb 20 21:56:01 2014
config:

	NAME        STATE     READ WRITE CKSUM
	zpool0      ONLINE       0     0     0
	  raidz1-0  ONLINE       0     0     0
	    ada0    ONLINE       0     0     0
	    ada1    ONLINE       0     0     0
	    ada2    ONLINE       0     0     0
	    ada3    ONLINE       0     0     0
	    ada4    ONLINE       0     0     4
	cache
	  ada4      ONLINE       0     0     0

errors: No known data errors
Any ideas how I can remove the cache drive there? I should probably also find a way to force NAS4Free to use UUIDs . . .

IOwnCalculus
Apr 2, 2003





That's... odd.

What shows up if you hit the shell and run 'zpool iostat -v'?

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply
«608 »