Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Locked thread
Smackbilly
Jan 3, 2001
What kind of a name is Pizza Organ! anyway?
I've just put together a computer whose sole purpose is to take rsnapshot backups of my other three computers. Since rsnapshot is incremental, and the computer has no other purpose, the drives are only active for a few minutes at a time (when the backups occur). So I've thought about setting the drives to spin down when they are not needed in order to conserve power, and to reduce heat and noise generated.

Currently, I have the backup process set to run every hour (to take advantage of rsnapshot's ability to work as a sort of rollback service). So what happens is that the drives spin up, do I/O for 5-15 minutes, and then spin down. I've controlled the other aspects of the computer (logging, etc) so that the drives don't spin up except when the backups are being made or accessed.

My question is: is it significantly worse to spin the hard drives up and down once an hour as opposed to leaving them on constantly? I don't really care if it's going to shorten the lifetime of the drives somewhat, but if I'm going to burn them out in a matter of months, I'll have to rethink this.

Adbot
ADBOT LOVES YOU

Smackbilly
Jan 3, 2001
What kind of a name is Pizza Organ! anyway?

VERTiG0 posted:

Also what is Samba filesharing/SMB or whatever, and should I care?

The SMB protocol is the techno-jargon name for Windows file sharing and network authentication. Samba is a program for Linux and similar systems which allows a non-windows machine to participate in windows file sharing and/or act as a domain controller.

Smackbilly
Jan 3, 2001
What kind of a name is Pizza Organ! anyway?
What is the best way to combine disks into a single logical partition such that:

- The combination can tolerate at least one drive failure
- The size of the set is expandable, preferably to asymmetrically sized disks
- The ratio of usable space to total capacity is maximized

?

I've thought of two possible solutions, but neither seems all that great:

1. RAID 5 array.

Benefits: Can tolerate 1 failure, usable space is capacity - 1 drive

Problems: Cannot expand without erasing and rebuilding, disk sizes must be symmetric

2. LVM over RAID 1 arrays

Benefits: Can tolerate multiple failures, can expand without erasing and rebuilding, disk sizes must only be symmetric within a pair

Problems: Usable space is only half of capacity


The ultimate goal is to come up with a system wherein I can add disks at will to increase the size of my storage, and if I want to remove a small disk and replace it with a large disk, that's also not a problem. Straight LVM would do exactly this, however that's basically JBOD, and therefore any failure loses all my data. I don't want that, but on the other hand, this is for home use and I do backups of my important stuff, so dedicating a bunch of extra space to protecting against multiple failures is a waste of money - protecting against a single failure is sufficient for me.

Is there some way to get the flexibility of a JBOD/LVM solution while still being able to tolerate a single failure, that doesn't involve all the wasted space of LVM over RAID 1 arrays?

Smackbilly
Jan 3, 2001
What kind of a name is Pizza Organ! anyway?

Saukkis posted:

How about LVM over RAID5s? Most flexible system is probably one where the harddrives are split into small (~20GB) equal sized partitions that are RAID5ed and then these separate RAIDs are combined with LVM. When you want to add a disk you split it in partitions same way as the other drives. Then you use LVM to empty one of the RAIDS, recreate it with the new disk added and then you take this to use with LVM. Repeat for every RAID5. This is probably the system I'll start to use.

That sounds like an interesting idea, but I was hoping to use hardware RAID (yes, yes, I know motherboard-builtin hardware RAID sucks, but it's better than software RAID, right?). I was under the impression that the smallest unit involved in a hardware RAID would be a disk, not a partition. Is that wrong? Would a hardware RAID controller be able to do a RAID 5 across partitions?

If I understand this right, you're saying that with this method you will have an amount of space equal to one of your RAIDs that is basically a temporary area where data is moved while the RAIDs are expanded, and this area is entirely unused during normal operation? Which is why you'd want to keep your partitions small.

So if I did this with 3 300GB drives, each drive would be split into 15 20GB partitions, and joined into 15 different RAID 5 arrays, one of which would be unused except during expansion, so the unusable space would be 14 * 20GB for parity and 3 * 20G for expansion, so about 62% usable space, which improves to 83% usable space if I add a fourth drive of the same size, since the parity space doesn't increase.

The other issues I see here are:

1. If I'm going to add larger drives I have to do it in pairs, but I suppose that's no worse than LVM over RAID 1

2. Running 15 separate RAID arrays on one machine seems pretty absurd - are there any problems that would arise from doing something like this?

  • Locked thread