Комментарии:
A lot of people are having trouble understanding that yes, a single URE can indeed cause the entire array to fail to rebuild. Because of the nature of the parity system, it is NOT the same as having a URE on a standalone drive, where only that single file is lost. You need it ALL to rebuild it all. You might be able to send the drives to a data recovery service who could potentially rebuild the array and only lose the files directly using the lost parity data, but that would be very expensive and time consuming, and not necessarily a guarantee.
Not to mention that because array rebuilds take so long, and are very very intensive, it drastically increases the chances of having yet another URE in that time, which could completely ruin any chances of any recovery at all.
You can google something like "Raid 5 rebuild failure probability calculator" if you don't believe me on the odds. But using 10^14 URE rate drives is extremely risky and you actually are VERY likely to experience a rebuild failure even with small arrays.
In any case, when looking to buy drives for a NAS, my suggestion is ONLY buy a drive with a URE rate equal to or better than 1 per 10^15 bits. This can be found by googling for the 'Data Sheet' of the specific model drive you're looking for. It will be called something like "Non-recoverable read errors rate", "Error Rate (non-recoverable)", "Unrecoverable Bit Error Rate (UBER)", or something similar, probably listed along with the 'reliability' specs in the data sheet.
I haven't really been able to even find any HDDs with a URE rate better than 10^15. However, with SSDs you can usually find better URE rates. Data sheets for consumer desktop SSDs (Like samsung's EVO/PRO SSD line or Seagate's Barracuda SSD line) don't list URE rates. But Seagate's "Ironwolf" NAS SSD line actually have URE rates of 1 in 10^17, so 100 times better than any HDD I've seen. However those cost several times more than an HDD (right now it's $750 for just a 3.8TB drive).
Yeah raid is not backup. Had my first two disc fail in a raid 5 but backup saved the day......
ОтветитьWhat do you guys recommend me for a torrent movie library? (Don't judge me, where I live Blu rays are not sold and streaming quality is not on par for OLED TVs) I want to have around 50TBs in RAID5 or 6. As far as I know HGST is the best brand. Is it still the same today? And what NAS Brand?
ОтветитьThis video is waaaaaay wrong. The reason why you shouldn’t use RAID5 has nothing to do with any capacity loss. It has to do with the fact that in most RAID5 configurations there is only one parity check source. There’s no way to actually tell if a drive is lying. Enterprise raid controllers worked around this by using 520 byte per sector formatted hard drives, and they put a separate checksum on the extra 8 bytes. With RAID 6 this becomes unnecessary because you now have P and Q, two different stripes of parity data per I-block. You give up capacity but that extra parity stripe gives the raid engine the ability to make more intelligent decisions regarding data corruption. Also in regards to unrecoverable read errors, you can still get the data back so long as you can read from the original drive that puked. It will have a “stale” config but still be accessible and if that part of the volume wasn’t written to then you’re still good to go.
ОтветитьI'll stick with RAID 5 (4-6 drives) and nightly backups to LTO tape. My data is not mission critical as some, say like, in the financial sector where one bit can destroy the balance sheet.
ОтветитьHow does raid 5 or 6 increase write speed?
ОтветитьRAID 5 with good back up strategy is better than focusing too much on RAID.
ОтветитьA lot of newer harddrives like the ironwolf pro have one URE in 10^15 bits instead of 10^14. So that means 125 tb of storage where one bit will fail you.
ОтветитьRaid 6 has slower write speed than raid 5, because you need to write parity twice. Raid 50 is better.
ОтветитьIts funny: We want more parity (same data for restore) but not more duplicants (same data)
So on one side we want more RAID and on the other side we have de-duplication
Raid! Shadowlegends
ОтветитьNow I feel guilty for running Raid 5. Wishing I did 6.
Ответитьgreat video, but in the comment need to talk about newer RAID levels like z raid which i believe doesn't have the problem
Ответитьhow about home environment with a total of 3-4 drives?
ОтветитьChris Titus Tech teaches us RAID 6 is for newbies and shouldn’t even exist, and I’d have to agree with him RAID5 is better
ОтветитьIs all the data really lost though, with the read error? Wouldn't you have like one or two corrupt files, but still get the rest back?
Ответитьwindows server 2019 dosent support raid 6 how can you set it up thanks
Ответитьnot sure if you are still monitoring this feed, but any updates on RAID systems and/or tech on this subject. Dusting off an old project and want to finish it this winter. Thanks
ОтветитьDrives almost never fail where you can't do anything with them. 99% of the time its a slow death where data is still recoverable. When a drive has issues I would recommend taking it out of the array and copying data over to another disk in case. Then rebuilding the array
Ответитьfunny old thojoe
ОтветитьI lost a primary domain controller once back in 2003. The raid 5 didn't help, neither did the secondary domain controller. It was a huge mess. I had to accept defeat and rebuild everything from scratch in two weeks with the help of a friend. Slept at the office for that time. So now I am a bit skeptical to all redundancy systems :)
Ответитьand its compounded in that ALL of the drives have EVERY sector read during a rebuild for maximum heat and least reliability. With 8 * 18TB your talking about 24 hours non-stop to rebuild
ОтветитьIf you can sustain a temporary loss in time, or a slight loss in data, then you can get the best of both worlds with a RAID 0.
With RAID 0, you will get maximum performance. And to mitigate the loss of your entire storage system if even on drive fails, then do a backup.
If your RAID 0 fails, you simply rebuild the RAID (after you replace the drive that failed), and you then populate your rebuilt RAID 0 with data from your backup.
But if you are running a business, where you can neither withstand down-time nor any data loss, the scratch any ideas of using RAID 0.
Regardless of which RAID level you choose, or even if you are not using RAID, you should still be doing backups, and if downtime is a big deal, then have a spare drive at the ready, to replace a drive failure. Otherwise, you will have to order one and wait for delivery, or go to a store (assuming you have one, locally), drop what you are doing, and buy a replacement drive. It is no fun to have to deal with that, unexpectedly. So have a spare drive on hand.
In 16 years I've had two disk failures in RAID5 on two NAS's. Both recovered without any drama whatsover. You should also renew the disks every 5-6 years. Plus S.M.A.R.T. will warn you of impending issues.RAID5 was never intended as a data backup so I don't understand the criticism.
ОтветитьI still can't believe this guy grew up telling complete lies
ОтветитьSimplest view - most people who by "same model/size drives" often get from the same issue of drives or similar age. Drives in RAID of this sort often age about the same rate and probability goes up with a single failure that more are to follow. Does RAID 6 make sense? In one way as promoted in the video the answer is yes. However, aged drives really should be replaced and staggered in replacement. Well that might make sense except rebuilding each time creates a new risk. There are other RAID options where one might take a hit on total real storage available but may be less risky - consider RAID 10 which has good fault tolerance and can allow for 2 out of 4 drives to fail and so on. It too is not without risks.
I have a 5 bay NAS with 4 tb drives. It was used for about three years and put into cold storage. I recently took it out and had about 8 terabytes of data remaining on it. Lucky for me I did a deep scan and found the old drives not showing any errors. If I had the opportunity, I might have use 4 larger drives with raid 10 and add a hot spare later.
I'll simply say this has been a discussion for many many years on what should be done with the traditional RAID options.
another thing to take note is when a RAID 5 is rebuilding it works your existing(possibley near EOL) drives harder to re build the array. I have had 2 drives fail on me in the middle of a rebuild
ОтветитьI have one drive. can I use raid 7 ?
ОтветитьGreat information. Very useful video 👍.
In the future, I will look for (quoting you):
"Non-recoverable read errors rate", "Error Rate (non-recoverable)", "Unrecoverable Bit Error Rate (UBER)", or something similar, probably listed along with the 'reliability' specs in the data sheet.
Rated at 1e15
Would S.M.A.R.T give you any warning of impending URE?
ОтветитьWith RAID 5 and a hot spare don't you get the same result as RAID 6 except you have less probability for URE because you have less drives in the RAID?
Ответитьcan raid 5 be converted to raid 6 in windows 10 without losing data?
ОтветитьIf I only have 3 HDD I think RAID 6 becomes worthless.
You lose 2/3 of the space. At that point is the same as using RAID 1+0 with 2 partitions per drive.
Raid 10! And You forgot about all this things
ОтветитьRAID 6 factorial does not exist
ОтветитьQuestion: let's say if one drive fails in RAID5 array, can I just remove the bad drive and access the data, I mean, before adding a new drive and rebuilding the array?
If yes, then I can move all the data to another storage system. In this case, if one of the remaining drives has URE, then I guess I will just loose very few data but not all of them?
You are aware that you lose performance in a raid 6 array.
I work for an MSP that specializes in dental I.T. These places due to Canadian standards must keep data for patients for 10 years. Every one of our 2000 servers in play all have a raid 5 arrry and for the most part there are no issues. Now keep on mind even with a raid array we still run a cloud and on site back up of the host and the DC VM and Server VM. You should never %100 rely on you raid array.
RAID 6 will not follow the fate of RAID 5 since when you loose one drive, 2 UREs at the same time (in the same segment) are highly unlikely. You can have 1 URE in every segment and still be able to recover it.
ОтветитьWhat about using only 3 drives for RAID 5?
ОтветитьRAID 60 is the way.
ОтветитьUse raid 10
ОтветитьAnd here is a person who consists 12TB x 4 for Raid 5.
Ответитьthink server rd 440 sas hdd frequently faulty
ОтветитьRaid 1 is only usfull iff you are using seagate
ОтветитьThanks for the education. If I'm using SSD drives and do very little writing (mostly just archives) on a 12TB array (2TB x 6) Should I go for RAID 6 or SHR2? or just go with SHR or RAID 5?
Ответитьthis video is super helpful, thank you for making such a clean and straight video!
Ответить