Insert stupid pun about raiding things here
-
I'm about to be in possession of 3x WD 3TB Red hard drivers.
Can I do something like this:
- Stuff the first hard drive into a box
- Jam a RAID controller in there, too
- Mash the second hdd into the same box
- Tell RAID "yo yo yo these are one in the same make them safe"
- Plow that third drive into, let's say, an external USB dock
- Tell the whole system "mirror your shit to that dock"
When the system is done mirroring, take the third hard drive out of the dock, and offsite. Bring it back about once a week to update it.
OR is it better to buy fourth hard drive and rotate the "offsite" drive? Or something else?
-
@Lorne-Kates said in Insert stupid pun about raiding things here:
Plow that third drive into, let's say, an external USB dock
Leeroy Jenkins!
-
@Lorne-Kates How much of the system has to be duplicated?
Would it make more sense to host a system offsite and use ftp?
-
@Lorne-Kates
A single 2 drive mirror (RAID 1) would theoretically allow your data to be preserved in the case of a single drive failure. However, most people that do RAID 1 often do so with the cheap on-motherboard "RAID controllers", and usually see failures such as one drive partially fails, and then the RAID controller gets confused and tries to rebuild the array from the corrupted drive to the good drive, making the whole thing go poof. (Usually occurs on reboot after a power failure or other dirty shutdown).Also, for the third drive, I would recommend using backup software to effect the data transfer from your online array. trying to do some sort of direct mirror (I inferred an additional software RAID mirror from your phrasing) is a pretty big headache at best -- RAID isn't designed for an offline backup drive, and there's a significant possibility that your backup drive would be the "rebuild source" and overwrite your production drive when you reconnected it to resync. I strongly recommend using a purpose-built solution for your offline backup.
And, as a general reminder - RAID isn't backup. It will happily propagate your ransom ware installation, file deletions, saves of incorrect data, and spontaneous decision to purge that rare pron to keep the kiddo safe. And once you do it, there's no undo button from the RAID. You do still need a backup in addition (especially for critical business data or photo albums!)
-
Echoing what @izzion said: Raid isn't backup, don't think of it like one. It's more of an incidental first-line-of-defense that can keep you going if something breaks.
Personally, I use FreeNAS, which is based on FreeBSD and uses the ZFS filesystem which (along with RAID capability) allows things like incremental snapshot of the filesystem, streaming backups, and a slew of other things.
I went with a 4-drive setup, though 3 drives would still fit in with a fairly redundant configuration that would still allow recovery from @izzion's disaster example.
Of course, FreeNAS isn't easy on the specs it needs to run, the latest releases need a 64-bit CPU and (bare minimum) 8 GB of ECC Ram (more if you can handle it, the in-memory cache is REALLY hungry), and about 16 Gb for the OS disk (Flash media was recommended at one point because once the OS is loaded it doesn't really need to access it a whole lot).
Once you've got your foundation, as @izzion said, get a real backup solution. FreeNAS's ZFS itself can be replicated to other ZFS-capable targets (such as another FreeNAS/FreeBSD machine) and there are a few providers that let you transmit to them that way. But there's also third-party tools like CrashPlan and the like that should do the job just as well, and would offer restoration on a more granular level than "Well, please wait while we rebuild the entire hard drive before you can access it".
Technically speaking though, yes, your sketch of a solution could work, but it's very susceptible to user error and the practices you would be implementing would likely stress the disks more than needed and possibly provoke early failure in a drive, and it's somewhat common that when one drive fails, the stress of rebuilding a new one can cause one of the remaining drives to fail (that's when life REALLY gets crappy, it is not fun whatsoever...).
-
It seems nobody listed the risk of having your RAID controller fried. Though a RAID 1 drive could in theory be connected directly to a system, it is prudent to assume that it is worthless when you connect it to a different RAID or HD controller.
-
@JBert said in Insert stupid pun about raiding things here:
It seems nobody listed the risk of having your RAID controller fried. Though a RAID 1 drive could in theory be connected directly to a system, it is prudent to assume that it is worthless when you connect it to a different RAID or HD controller.
That's one of the nice things of having the raid information stored on the disks themselves (like ZFS), it's no longer dependent on the controller to figure out which drive is what.
-
how long until someone tells @Lorne-Kates that mirror is not the same as backup?
-
@Helix said in Insert stupid pun about raiding things here:
how long until someone tells @Lorne-Kates that mirror is not the same as backup?
Soon!
-
@Lorne-Kates said in Insert stupid pun about raiding things here:
Or something else?
Get two NAS containers and at least one more drive. Configure the two NAS containers so that one is your working storage and the other is the backup system. Use a nightly incremental backup scheme. (At least some NAS systems come with reasonable backup software when you're backing up to another NAS by the same manufacturer.) Job done, doesn't cost too much, and its both correct and simple to keep working.
Amount of storage and number of replications will depend on your budget.
-
I've got a CentOS 6 with ZFS raidz2 (basically raid 6) and backing up the whole thing to crashplan.
Now to get a cheap crashplan renewal. Anybody know where?
-
@izzion said in Insert stupid pun about raiding things here:
hen the RAID controller gets confused and tries to rebuild the array from the corrupted drive to the good drive, making the whole thing go poof.
Been there, done that. RAID-1, in my experience, can't even do the one thing it's supposed to do.
-
@mott555 said in Insert stupid pun about raiding things here:
RAID-1, in my experience, can't even do the one thing it's supposed to do.
Been using software RAID-1 on Linux for more than 10 years and it never failed me.
Cheapo onboard RAID controllers on the other hand are
-
@TimeBandit It happened to me on a home ESXi server built with consumer (but overclocker)-grade hardware. I don't think software RAID is even an option given the choice of operating system.
-
@mott555 said in Insert stupid pun about raiding things here:
@TimeBandit It happened to me on a home ESXi server built with consumer (but overclocker)-grade hardware. I don't think software RAID is even an option given the choice of operating system.
ESXi won't let you use software-RAID. Which is a shame, since it's based on Linux.
One of the advantage of software-RAID-1 (on Linux) is, you can take any of the 2 drives and mount it on another Linux machine, it's just a normal Ext-2/3/4 partition.
-
@TimeBandit
Toby Faire, most of the RAID-1 failures I've had to clean up after resulted from some sort of user error or neglect- Someone "accidentally" removed a drive from the RAID and immediately reinserted it, and the controller tried to rebuild using the removed drive as master and blew everything up
- The computer had been suffering the effects of dirty power and unexpected shutdowns for a while, the ID-10 Mark T in charge of the computer elected to ignore our recommendations to get a UPS, and eventually the RAID came up with one drive marked offline. He panicked, did something (never copped to exactly what) and triggered a rebuild operation that wiped both drives.
I've also had RAIDs of various types that suffered a single drive loss, then failed during the rebuild because a second drive burned out during the rebuild. Those are usually traceable to all the drives being from the same manufacturer and lot, so when one fails the others are usually pretty close to the end of useful life and throwing extreme usage at them with a rebuild will highlight whichever drive is weakest, often killing it. Not a RAID-1 specific issue, but another great example of how RAID Is Not Backup.
Edit: I also once had a very pricey HP SAN (really 2 SANs, you ordered them in pairs) with 12 drives in a RAID 50 format (6 drives RAID 5 + 6 drives RAID 5 striped together) then mirrored to a second SAN. So basically a 24 drive RAID 501 that got you 10 useable drives worth of space but very good performance and redundancy. We had one drive in those trip its SMART flag and start clamoring for replacement (thankfully while we were still under support/warranty). HP Support had a diagnostic for us to run before they shipped the replacement drive, it highlighted 5 more drives that were near failure, they shipped us all 6 drives and a procedure to follow to rebuild without data loss. Three weeks later, we had finally finished replacing a total of 10 drives and effectively rebuilding the one unit from the other (most of the bad disks were concentrated in one of the two units), but we had no downtime or data loss through the process. Still a serious PITA though.
-
@izzion said in Insert stupid pun about raiding things here:
I've also had RAIDs of various types that suffered a single drive loss, then failed during the rebuild because a second drive burned out during the rebuild. Those are usually traceable to all the drives being from the same manufacturer and lot, so when one fails the others are usually pretty close to the end of useful life and throwing extreme usage at them with a rebuild will highlight whichever drive is weakest, often killing it.
That's why I always build my RAID from 2 drives bought from 2 different places, to make sure they are not from the same batch.
another great example of how RAID Is Not Backup.
I agree
-
I use RAID0 (at home) for speed. Most of the data can be replaced, and the important stuff that can't is duplicated into a cloud somewhere.
-
@xaade said in Insert stupid pun about raiding things here:
Would it make more sense to host a system offsite and use ftp?
Might as well save the cost of hosting another system and use something cloudy if it's an offsite copy.
-
@loopback0 said in Insert stupid pun about raiding things here:
the important stuff that can't is duplicated into a cloud somewhere
You'd better hope it doesn't rain.
-
@RaceProUK Cloud insurance.
-
@loopback0 said in Insert stupid pun about raiding things here:
Cloud insurance
Shouldn't be a surprise:
-
@TimeBandit said in Insert stupid pun about raiding things here:
That's why I always build my RAID from 2 drives bought from 2 different places, to make sure they are not from the same batch.
That's a good call. To add to the anecdotes, I have also personally had a second drive go out on a RAID 5, even though I hot swapped the first drive immediately when an SNMP trap went out for its failure. That needle should have been impossible to thread.
Sad day for the primary domain controller. :<
-
@heterodox Rebuild is a heavy wear on a drive.
-
@PleegWat said in Insert stupid pun about raiding things here:
@heterodox Rebuild is a heavy wear on a drive.
I grant that, but drives that were sold to us in a RAID 5 configuration from the OEM should have been able to handle it. In any case, just adding my lovely voice to the "RAID is not backup" chorus.
-
@heterodox said in Insert stupid pun about raiding things here:
I have also personally had a second drive go out on a RAID 5, even though I hot swapped the first drive immediately when an SNMP trap went out for its failure.
That's why RAID is not a substitute to proper backup
-
I heard RAID isn't a backup.
-
@TimeBandit said in Insert stupid pun about raiding things here:
That's why RAID is not a substitute to proper backup
That's what I'm fucking-- yes.
-
@loopback0
I'm reminded of the old saw about RAID 0:"RAID 0 is the only one that's honest, it tells you exactly how much of your data you'll have left the first time something abnormal happens."
-
@heterodox said in Insert stupid pun about raiding things here:
@TimeBandit said in Insert stupid pun about raiding things here:
That's why RAID is not a substitute to proper backup
That's what I'm fucking-- yes.
I think I see why your hard drives keep crashing.
Filed under: Solid State
-
This thread made me think the only RAID that makes sense is RAID 0.
-
@Gąska I like Raid 1 for the improved iops on random read. I do my backups on Raid 0 of course because it's got the best write performance.
-
@gleemonk is the difference in random read between 1 and 0 noticable?
-
@Gąska said in Insert stupid pun about raiding things here:
This thread made me think the only RAID that makes sense is RAID 0.
it depends HEAVILY on what you actually want to do with it mind.
Raid 0 - Optimizes Performance at the cost of resiliency (if either drive fails you lost everything)
Raid 1 - Optimizes Resiliency at the cost of performance (though not that much to be honest... it is noticable though)
Raid Other - try to balance resiliency with cost effectiveness. it's expensive to have 2x or 3x the number of disks you need for just data, so if you're willing to give up a bit of resiliency (each dish isn't cloneable in isolation) you can have decent resiliency with only 10-20% extra disks. (depending on setup)
Raid Specialized - if you need this, not only do you know it but you know what they are for.and of course.... RAID AIN'T BACKUP. if it was backup there's be a
B
in the acronym. and i can't find one there.
-
@Gąska I don't actually know because I never tried Raid 0. Maybe I should have said "If I did backups I would do them on Raid 0."
On a more serious note the two systems I built on Raid 1 did seem to stall less, I didn't measure though. And that was right around the time when 30GB SSD became affordable at which point I lost interest in Raid.
-
@gleemonk oh right, SSD. Forgot about them.
Man, I feel old... and I'm not even out of college!
-
@Gąska said in Insert stupid pun about raiding things here:
Man, I feel old... and I'm not even out of college!
Only goes downhill from where you are. :P
-
@accalia said in Insert stupid pun about raiding things here:
Raid 1 - Optimizes Resiliency at the cost of performance (though not that much to be honest... it is noticable though)
More precisely, you get a small performance drop on write (have to write it twice) and gain a small performance on read (you spread the read on both drives at the same time).
-
@Gąska
Well, I mean, I have relatives that are in their 40s (or better) and are not out or recently out of college. So it's entirely possible you may, in fact, be old.
-
@loopback0 said in Insert stupid pun about raiding things here:
I heard RAID isn't a backup.
Right. You're supposed to use git, I think.
-
@boomzilla RAIG?
-
@RaceProUK works well for Linux.
-
Hey guys, I can't afford a tape drive and disks are so cheap, I was thinking about setting up a RAID for my backups. What's the best way to configure it? I just want to back up some photos.
-
@another_sam said in Insert stupid pun about raiding things here:
Hey guys, I can't afford a tape drive and disks are so cheap
Tapes are more expensive than disks?!
Oh, you're not talking like audio cassette, are you?
-
@another_sam
It's easier just to call the FBI if you lose any of your special photos. They already have a copy preserved for you.
-
@Tsaukpaetra said in Insert stupid pun about raiding things here:
Tapes are more expensive than disks?!
Tape drives are, very much so.
@Tsaukpaetra said in Insert stupid pun about raiding things here:
Oh, you're not talking like audio cassette, are you?
Only us 40-year-old Atari and Commodore users know what you're talking about.
-
@another_sam said in Insert stupid pun about raiding things here:
@Tsaukpaetra said in Insert stupid pun about raiding things here:
Tapes are more expensive than disks?!
Tape drives are, very much so.
@Tsaukpaetra said in Insert stupid pun about raiding things here:
Oh, you're not talking like audio cassette, are you?
Only us 40-year-old Atari and Commodore users know what you're talking about.
I've always wanted to play with a multi-modem setup that would write out several datastreams across several sets of frequencies to tape, see if that would help the data density while still being performant.
Never got the time though...
-
@another_sam said in Insert stupid pun about raiding things here:
Tape drives are
That reminds me of the episode of Burn Notice where Michael was running all over town trying to find someone who could read the "tape drive" the bad guy du jour had left behind for our hero to find. I wanted to punch my TV so many times during that episode.
Spoiler:
If they had left a tape DRIVE, you would either be gold for reading it or have nothing since you didn't have the tape! But all you have is the TAPE, and now you need someone with the right tape drive! Auuuughwharrrgarbl!!!!!!!
-
@Tsaukpaetra said in Insert stupid pun about raiding things here:
I've always wanted to play with a multi-modem setup that would write out several datastreams across several sets of frequencies to tape, see if that would help the data density while still being performant.
You have strange hobbies.
-
@another_sam said in Insert stupid pun about raiding things here:
@Tsaukpaetra said in Insert stupid pun about raiding things here:
I've always wanted to play with a multi-modem setup that would write out several datastreams across several sets of frequencies to tape, see if that would help the data density while still being performant.
You have strange hobbies.
Are they really hobbies if they never get past the thought-experiment stage?