[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Homebuilt NAS Advice



On 8/6/2020 11:08 AM, Stefan Monnier wrote:
	For better performance, more space, and higher throughput, I would
	probably create a RAID 4 or RAID 6 array from the external enclosure
	and use it as the data repository.

And you suggest I put a 4-drive enclosure in my backpack next to my laptop?
Seriously?

	Yes.  Also a bottle of water and a pocket knife.

For the BananaPi, the suggestion is marginally less problematic but
still: a non-trivial constraint with significant immediate downsides.
	Such as?

Extra space taken, extra power used 24/7 (which in turn requires an
extra plug because the poor BananaPi can't provide all that power),

	Now it is my turn to ask, "Seriously?"

extra costs (the single drive I use was itself more than 3/4 the price
of the whole system, so adding a second drive (complete with external
power supply and enclosure) would double the price of the system),

The Banana Pi is indeed extremely economical. I have ten of them, plus 17 Raspberry Pi 3B computers.

Extra failures (more hardware => more failures), ...

	More failures on average, but far less serious and costly ones.


RAID is basically an insurance.
	Not entirely.  A RAID 5 or RAID 6 array is far, far faster than
	a single hard drive.

Right, RAID-over-USB is of course going to blow my SSD-over-SATA out of
the water by a wide margin (not!).

On the Banana Pi it can. On the laptop, probably not, assuming as you say, an SSD.

	It is also much larger than a single hard drive, sometimes at
	less expense than a single large hard drive.

That's great if you happen to be in that spot, but that's not my case.

What spot? You mentioned cost. For many configurations, 4 small drives are cheaper than 1 large one.


	It is also portable from one system to another.  Unplug the array
	from the laptop, plug it in to the Banana Pi, and presto! The array
	is now attached to the Pi.

Wonderful.  But then it's not a RAID shared with the internal drive
any more.  So it won't protect my root partition.

It can, but I probably would not. Assuimng / is, say, 20G or so, one can create a loop device on the array of the same size and make the internal drive and the loop device a RAID 1 mirror. If the main drive fails, install a new drive, partition it, boot from a thumb drive, make the new partition a mirror of the eternal file. Go away for an hour or two.

The same can be done making a partition on the external drives instead of a loop device, but it is a little less flexible.

And if I keep my home
partition in it, the "presto" comes with the footnote "after you logout
and log back in" (fun!)

	You lost me.  Why log out?

but if I don't keep my home partition on it,
then my home partition is again not protects by RAID at which point I'm
starting to wonder what I would put on that RAID.

Data, but then I definitely recommend putting /home on the external array, so the question is a bit moot.

	It's really not any different logically than an external drive,
	except it is faster, larger, and more robust.

It's no different, indeed, except a bit more expensive and bigger.

Well, OK. How much is down-time worth? If you consider the cost of the downtime associated with a failed system to be trivial, then that aspect is not important.

But more importantly: there's a reason why I'm not using an external
drive at all in the first place.

I am sure there is. Do you admit that external drives are extremely popular? Literally millions of them are sold every year.

If a disk dies in the RAID, you still have to replace it

Well, one should. It's not required, and there have been cases where I could not replace a failed drive for several days.

and the
performance is affected for a while.  The cost is never
completely eliminated.

It is affected during the rebuild, not the time between the failure and replacing the drive. One can manage the amount of impact, however, and one can always delay the rebuild until off hours.


anyway (not to mention that RAID doesn't prevent me from losing work
when the OS or my editor crashes
	Uh, yeah, it can.  It definitely can be used to eliminate data loss
	when the OS crashes,

Of course it can't: when the OS crashes, the unsaved work in your editor
is lost.

Yes, of course. I mistook what you were saying. Are you suggesting that is not the case for a non-RAID system, however?

	That is another matter.  Indeed, it is probably the most likely
	reason a need for a backup solution exists.

For my use-case, RAID would cost a fair bit of money and inconvenience,
and the benefit would be rare and minor.

How much does downtime and time to rebuild a system cost you? For many of us, even an hour of downtime is very expensive, and rebuilding a machine during peak hours can be hideously costly. It is always a pain in the ass.

I suspect your "world" is one where you have U1 blades and things like

No. My servers and workstations are relatively inexpensive desktop class clones.

that and/or where the disk's content is not 100% your own (i.e. other
people depend on it)

Not exactly. It's not the content upo which people rely, per se, but the ability of the systems to remain online. More importantly in my case, *I* rely on the systems to be up continuously. Downtime can be very expensive for me. Certainly more than the cost of a hard drive.

so for your use case RAID is probably
a no-brainer.  For me, the choice is similarly blindingly obvious but
diametrically opposed because my use case is radically different.

Perhaps so. How many people - especially Debian users - consider a catastrophic failure to be cost-less? Debian's paradigm is based upon being rock solid and stable. Ubuntu gives up stability for the sake of new bells and whistles. I expect the typical Debian user is concerned about things like downtime and reliability. You seem not to be. If you are aware of and willing to take the risk, then that is fine. I certainly am not, and as a responsible professional I certainly would not ever recommend it. If anyone loses something based on my recommendation, they are very likely to come back and scream at me. I really don't care for that to happen.

Safety at all levels is always my top consideration. The second is safety, followed by safety. After that, it is speed, efficiency, and cost, in that order, unless they impact safety. One could consider me obsessive, I suppose, but I cannot count the number of times being exceedingly cautious has saved my butt, and even my life. Unfortunately, I also cannot count the number of times not being quite cautious enough has jumped up and bitten me hard in the butt.


Reply to: