[kwlug-disc] Bundling HD's. LVM vs mdadm vs ???

unsolicited unsolicited at swiz.ca
Fri Apr 12 02:46:52 EDT 2013


Interesting, thanks for the note.

> but wouldn't you prefer if your 2TB+ of data could survive a
> single-drive failure?

It already can. But thanks for the thought. Being a backup, it receives
the local data of a computer nightly, and itself replicates
elsewhere nightly. (via push/pulls, rsync/robocopy, as appropriate).

Which is not to say the suggestion of zfs isn't terrific - there appears
to be a lot of robust features there. Once set up, looks like I can zig
or zag as necessary, going forward.

> And it's WAF approved.

Got to be a story there. (-:
- I'm guessing you have a standalone box under the stairs or something.
Out of sight / mind / ugliness for WAF, while you get to sleep
peacefully at night.

Care to do a quickie presentation some kwlug night?
- surely the idea of robustly combining multiple disks into a single
presentation will be of interest to many.

On 13-04-11 10:17 AM, Chris Irwin wrote:
> I put my trust behind on ext4 on LVM on a mdadm RAID 1 mirror. Every
> distro supports those (though optionally, in Ubuntu's case), so your
> recovery options are open.
>
> Using something like ZFS (or eventually, BTRFS) would be nice, if
> only for checksums to verify the integrity of your data, which we
> don't get now. That said, I'd never rely on a file system I have to
> patch in or build myself. I want smarter people than me to be
> worrying about that part.

I hear you, but ... If zfs is a package, as are other filesystems (e.g.
jfs), and there are kernel updates (ext4), isn't every Linux
installation something you have to patch and build yourself?


If I'm understanding Chris' and Lori's messages correctly, from a user
interface perspective, LVM and zfs provide a front end / 'equivalence'
to each other, managing the magic underneath (user interface wise). e.g.
LVM does the mdadm handling.

It feels like LVM will grow to be / be replaced by zfs, eventually. Am I
misreading that?


On 13-04-11 09:17 AM, Khalid Baheyeldin wrote:
>
> Is 3TB enough? Because there are reasonablly priced 3TB disks out
> there.  I got two WD Green ones for backups.

It has been reported in this list that WD Greens have been problematic /
sleep issues. (I get you're popping a drive in and triggering a laptop
backup, but these are 24x7 systems with automated replications happening
every night. I prefer fire and forget. 24x7 as in on, not as in being
constantly beaten to death 24 hours a day - although, given the number 
of backups flying about each night, that's debatable. )

Since space requirements only ever go up, and cases only hold so many
drives (without additional expense), it seems prudent to be going with
the largest capacity drives available at the time. (I bought 4 2TB
drives not too many years ago, and now ... I'm looking for bigger ones
already!)

>
> The trick is to use parted, with some trick that I forgot now (but
> easily Googleable), because fdisk does see over 2TB.

2TB is still a limit anywhere in Linux these days!!!


> If you want 2TB + 2TB = 4TB, then there are several options:
>
> 1. mdadm (raid0) -- it appears as /dev/md0, so you put whatever
> filesystem on top. 2. btrfs -- slooow compared to raid+ext4
>
> Either solution has the same problem... lose 1 disk, you lose the
> entire filesystem.

True as standalone too, for that matter. i.e. /dev/sd#

Being replicated backups, I'm not worried about losing one drive. There 
will be one or two others around containing everything. I'll have time 
to acquire a replacement and replicate it back. (I would worry if the 
loss of 1 drive killed the other of a set - I can live with the one file 
that crosses the disk boundaries being lost.) As long as they all don't 
die at once ...

Some years back I contacted the fire department and asked some 
questions. I was still archiving off to DVD's, and no fireproof storage 
seemed large enough to contain them all, at any reasonable price point. 
Some interesting points came out:
- VERY seldom is it that an entire residence disappears, the fire 
department will get there fast enough. Computers with backups at top and 
bottom of house ... one or both are pretty likely to survive.
- put things in closets: less likely for floor above to come down on it 
and extensively damage anything, nor fall down a floor itself.
- fireproof anything probably not worth the expense, the temperature 
probably won't get high enough / long enough for combustion. Even paper 
may be smoke or water damaged, it will still be readable. (Security is a 
different issue, and not that I wouldn't recommend fire rated filing 
cabinets.)
- put things in metal filing cabinets: less likely for smoke or water to 
gum up the works.
- they pull the power first, it's not water that kills things, it's the 
combination of water with power. (Corrosion.) Things will quite probably 
dry out OK.
- smoke won't in itself generally kill things, although it may reduce 
lifespan. Even something smoke infested (if it's significant inside a 
computer case even) should last long enough to get a copy off.
- typically it's not the fire that does the damage - it does not last 
long enough, or extensive enough, to prevent retrieval. It's the 
measures taken to control the fire that do you in.

- I think we're more at risk from user error than anything technical or 
physical. (rm -f /mnt/sda2/*, anyone?)

Had a pipe leak a couple years back - everything but the power bars (due 
to combination of water and power) survived - only because the drives 
and guts are in the middle of the non-desktop computer cases - the 
bottom of the computer cases got a little wet, but all turned out well. 
(All such now sit on top of a couple of 2x4's!) [Not convinced the same 
would be true of an external drive, or NAS. Keep things off the floor, 
people!]

My only real concern is if someone comes in and steals everything (all 
replicas) at the same time! (Offline / IP backups of so much detritus 
feels unpalatable.)



On 13-04-11 09:24 AM, L.D. Paniak wrote:
> You might want to consider ZFS on Linux: http://zfsonlinux.org/
>
> Among other things, ZFS replaces the RAID/LVM/filesystem model with
> a disk pool/filesystem model that is considerably more flexible and
> easy to expand.  Throw in snapshots, data scrubbing, block devices
> and other features and you have a free storage environment that is
> difficult to beat at any price. ZFS  raidz (~RAID5) is going to
> require more hardware than the JBOD configuration you suggest below,
>  but wouldn't you prefer if your 2TB+ of data could survive a
> single-drive failure?
>
> And it's WAF approved.
>
> On 04/11/2013 09:01 AM, unsolicited wrote:
>> I'm running out of backup space, so I'm probably going to buy a
>> larger drive, and move a current drive into another computer -
>> expecting to bundle it with the drive already there into a single
>> logical presentation. (Add 4TB, move 2TB -> 2TB + 2TB = 4TB.)
>>
>> I don't 'do' RAID (at least at home, where high availability isn't
>>  an issue), and mistrust striping - lose 1 drive, both be gone. I
>> prefer to say 'this' amount of space for 'this' purpose, so when I
>>  run out of space I'm aware of and address the issue.
>>
>> (Presumably a worst case scenario for such bundling upon failure
>> is losing one file - the one that crosses physical drives. Correct,
>>  even today?)
>>
>> I've known for some time LVM is out there, just never needed it.
>> There's also mdadm ('LINEAR') [Which I gather LVM calls upon.]
>>
>> Any particular pros / cons / preferences / suggestions / thoughts
>> for bundling multiple drives into a single logical drive?
>>
>> -----
>>
>> I'm guessing the only reasonable drive out there at the moment is a
>> WD Black - (5 yr warranty, 7200, SATA3, availability / stock
>> levels). Counter suggestions?



More information about the kwlug-disc mailing list