Article: Why RAID 5 stops working in 2009


bubbaQ

Recommended Posts

Multiple logical arrays is what I really want.  That way I can have movies, music, and backup arrays all operating from independent parity.  The total number of drives is not my most critical item.  Sure, I can build multiple servers, but that would cost far more.

 

HOWEVER, I like this news.  I may go to the "unraid6" even with my current set of drives.

 

 

Bill

Link to comment

I could definitely see value in being able to install a couple more drives in an unRAID box.  I likely would never hit 30, but could easily see hitting 18 or 20, maybe even 22.  (As Joe mentions, at some point the parity builds / checks will get too long and the ability to simulate a failed disk too slow to be of value.)

 

A second parity drive also seems like a nice feature for unRAID.  It would certainly distinguish unRAID from other tools in this class.

 

but ...

 

the ways that users loose data has had little to do with having 2 drives fail at the same time.  The problem is having accurate parity and then keeping it accurate through the problem and problem isolation phases to the point that you can actually use it to reconstruct.  I don't think that having two parity drives would make this any better.

 

This post highlights some common ways that users can lose data.  I think that unRAID could protect users against many of these situations with relatively little effort.  Although not as flashy as the 2nd parity drive, my guess is doing these things would save more data.

 

In particular I think it is important that users be allowed to perform a parity check without updating parity.  It would be really nice if, at the end, unRAID would ask the user if they wanted to apply parity corrections.  If the answer is yes, unRAID could recheck the sync error sectors and, if they are still wrong, correct them.  This would allow users to safely run a parity check test without risking contaminating parity in the process.

Link to comment
  • 1 month later...

Just thought I'd add in my 2c worth..

 

I work in one of the most major DC's in Australia and I can tell you, there is a couple of servers there that use only SATA disks and mostly all of them SCSI. From all these servers, the SATA disks servers are almost guaranteed to have 2 failed drives within 24 hours of each other (I'm yet to see just "one" drive failure). Naturally, the drives are replaced within 24 hours of each other and you got it, the next failure has 2 disks in a row.

 

Seeing Tom's earlier post gives me quite a bit of confidence that I'm going to be able to increase redundancy even more - I only have to stop the thieves from breaking into my house and stealing my server now.

 

Cheers

 

Link to comment
I only have to stop the thieves from breaking into my house and stealing my server now.

 

This is a bigger hurdle to unRAID adoption that you may know.

 

I do a fair amount of consulting on network architecture and design, often for medial and legal clients.  HIPA and RPC require client/patient data to be encrypted, to prevent access in case of theft or other breach.  That's fairly simple with TrueCrypt on a conventional Linux install, and it is practically transparent to the user.  It is much more complex to do it with unRAID.

Link to comment
  • 10 years later...
On 12/10/2008 at 3:20 PM, bubbaQ said:

 

This is a bigger hurdle to unRAID adoption that you may know.

 

I do a fair amount of consulting on network architecture and design, often for medial and legal clients.  HIPA and RPC require client/patient data to be encrypted, to prevent access in case of theft or other breach.  That's fairly simple with TrueCrypt on a conventional Linux install, and it is practically transparent to the user.  It is much more complex to do it with unRAID.

Most family homes have really smart places to hide a server, as long as the case isn't insanely large. You should always pick the coldest areas of the house, for some that's the North side under the roof (on the attick part that's beyond the isolation shields), but it can also be a barn outdoors, depending if it's easy to hide some cat7 cables towards the unraid server. I've decided to include the server as part of our home's ventilation system, i.e. it's where coldest air of the entire ventilation system goes in at the front of the case and the warmest air goes out, and since the fans make the exact same kind of noise, nobody would find this server ;-)

Link to comment
1 hour ago, fluisterben said:

Most family homes have really smart places to hide a server, as long as the case isn't insanely large. You should always pick the coldest areas of the house, for some that's the North side under the roof (on the attick part that's beyond the isolation shields), but it can also be a barn outdoors, depending if it's easy to hide some cat7 cables towards the unraid server. I've decided to include the server as part of our home's ventilation system, i.e. it's where coldest air of the entire ventilation system goes in at the front of the case and the warmest air goes out, and since the fans make the exact same kind of noise, nobody would find this server ;-)

congratulations on bumping up a 10+ year old thread.

 

 

Link to comment
1 hour ago, fluisterben said:

Nobody cared or complained about waking up older still current subjects on usenet, ever. Who invented this strange "stuff we discuss must be temporary" thing anyway, it's ridiculously silly.

this isn't usenet. 

Link to comment
3 hours ago, fluisterben said:

Who invented this strange "stuff we discuss must be temporary" thing anyway, it's ridiculously silly.

I think maybe common sense invented it. In real life, do you suddenly quote something someone said 10+ years ago and expect them to continue the conversation? 

 

1. People barely remember what they said last week and why, let alone 10 years ago..

2. Online this is even more true. Users could have moved on to other things in 10 years, they probably don't even use the forum anymore.

3. It's like expecting to get support on a 10+year old version of some software which has now evolved to something completely different in those 10 years.

Edited by strike
Link to comment
57 minutes ago, strike said:

I think maybe common sense invented it. In real life, do you suddenly quote something someone said 10+ years ago and expect them to continue the conversation?

 

Actually, I do. And it doesn't apply to internet fora, since all info can be re-read, so there is no 'memory loss' or anything of that nature. Again, I've been using usenet ever since The Backbone Cabal in the late 80s through our Universities, and I've seen/taken part in literally hundreds of old threads, sometimes over 10 years old (like this one here) that suddenly became current again. Look it up, usenet has never had any issues or need for forced sustainability constraints, and none of your arguments hold true or make any sense to me. And even if they would be valid, why is it bad to add to historic subjects later on? Perhaps knowledge has evolved and subjects can be answered better than 10+ years earlier. I saw an unanswered reply here, which was why I responded.

Link to comment
1 hour ago, strike said:

do you suddenly quote something someone said 10+ years ago and expect them to continue the conversation? 

I don't know about you, but my wife does this all the time (usually in anger) and expects me to remember everything...

  • Like 3
  • Haha 1
Link to comment
10 minutes ago, BRiT said:

But at least you get benefits from having to relive all those moments

What planet do you live on?  No benefits that I can think of, other than the brain exercise in trying to figure out what the hell actually happened 10 years ago... (and we're way past "make up s...x");

Edited by Squid
Link to comment
39 minutes ago, fluisterben said:

Actually, I do. And it doesn't apply to internet fora, since all info can be re-read, so there is no 'memory loss' or anything of that nature. Again, I've been using usenet ever since The Backbone Cabal in the late 80s through our Universities, and I've seen/taken part in literally hundreds of old threads, sometimes over 10 years old (like this one here) that suddenly became current again. Look it up, usenet has never had any issues or need for forced sustainability constraints, and none of your arguments hold true or make any sense to me. And even if they would be valid, why is it bad to add to historic subjects later on? Perhaps knowledge has evolved and subjects can be answered better than 10+ years earlier. I saw an unanswered reply here, which was why I responded.

 

this....is...the...lamest, most made up excuse for getting called out for posting on a 10+ year old thread.... in the history of the internet... all the way back to the Backbone Cabal days in the late 80's.....EVER! And you probably have bent yourself into believing it as justification too.

 

ALSO: don't reply to me calling you out now. Wait 10+ years, and let's all test your theory. I'd like to see how knowledge "has evolved" in your reply, and perhaps in 10+ years, you'll have a better answer.

 

  • Haha 2
Link to comment
33 minutes ago, Squid said:

I don't know about you, but my wife does this all the time (usually in anger) and expects me to remember everything...

LOL 😂

 

1 hour ago, fluisterben said:

Actually, I do

@Squid You didn't say your wife was on here too.... You could have introduced her sooner.. :P

Link to comment
2 hours ago, Squid said:

I don't know about you, but my wife does this all the time (usually in anger) and expects me to remember everything...

Oh yeah, and it always starts with the words

 

"Can we talk about something....."

 

It might be phrased as a question, but it's not, it's very much a statement, those words make my blood run cold as soon as I hear them, as my mind starts going over what it is I may (or may not) have done that I was supposed to (or not supposed to).

 

Last week it was a whole conversation about my driving..........

 

 

 

 

6 weeks previously.

 

I mean I love my wife, but really?!

Link to comment
10 hours ago, bonienl said:

It seems you guys learned a lot about your wife the past 10 years (so did I) ☺️ 

Yeah.  Anything and everything that I've ever done in my life ever is fair game to her.  But god forbid I bring up something from yesterday without her losing it and complaining that it was in the past.

Link to comment
  • 1 year later...
On 5/27/2019 at 11:18 AM, Squid said:

Yeah.  Anything and everything that I've ever done in my life ever is fair game to her.  But god forbid I bring up something from yesterday without her losing it and complaining that it was in the past.

 

On 5/27/2019 at 12:59 AM, BRiT said:

In 10 years this thread will be old enough to drink even in the US.

 

On 5/26/2019 at 2:21 PM, strike said:

I think maybe common sense invented it. In real life, do you suddenly quote something someone said 10+ years ago and expect them to continue the conversation? 

Yes, I am an example of someone who came here 12 years later, hoping to find some answers on this topic.  In 2008 Limetech announced their intention to implement dual parity. Now 12 years later, I come to revisit Unraid, and I find that dual parity is already here, but for the life of me I can't find any useful documentation about it. (I googled a lot.)  And this thread is linked from the wiki.

 

So, let me give you a specific example of what's been bugging me for over 10 years now:

 

Suppose you have a two disk array -- one parity and one data.  And suppose that some cosmic particle has magically flipped just one single bit on one of the two disks.  Either one.  You do a parity check, you find the discrepancy, but with single parity, you have no way but to ASSUME that the parity disk is the wrong one, so with 50 percent probability you will be writing the wrong bit to the parity disk and all will look "good". 

 

Now, I was hoping that when the dual parity gets implemented, then the system will actually be able to KNOW which disk has the wrong bit, and not just blindly assume that the parity disk is wrong.  If Unraid is indeed doing this now, can somebody please point me to the discussion about all that?

 

Edited by Pourko
Link to comment

Others have already brought up this point (here or elsewhere), so why are you ignoring it for your cosmic example? The drive itself will know what data is wrong since the hardware supports checksums of sectors, so if a data bit is flipped and it doesn't change the checksum information, the drive will report it.

 

In addition to that, you may know what's wrong if the filesystem used supports checksums or you use a 3rd party plugin for checksums.

 

Mathematically it is impossible to know which disk is wrong with the algorithms used by single and dual parity protections.

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.