• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

linux raid server

novadude

[H]ard|Gawd
Joined
Jul 8, 2004
Messages
1,108
I've decided that I want a media/file server. Before I get too far though, I need some questions answered. Linux software raid 5 supports spin down of the raid drives and expanding the array (both adding new drives of the same size, and new larger drives?) correct? EDIT: how about RAID level migration? can I move 1 drive with data into an array?

Also, with the following hardware could I expect at least 60 MB/s over gigabit ethernet with files over a few MB's in size? this would be housing files mostly from RAW pictures to my DVD collection, but speed will be important for editing the pics/vids so I'd love it if the gigabit were the limit and not the raid

hardware: (I have all but the drives laying around with no other use for, so rather than selling I'll just use it)
Abit IP-35pro, Intel e2160/e5200, 4GB ddr2, 6x Seagate 7200.11 1tb's

I'll probably be throwing that in a lian-li A05b with a 3 in 2 bay adapter or an Antec 300, a ~330w PSU and boot it from a CF->IDE adapter
 
Linux software raid 5 supports spin down of the raid drives and expanding the array (both adding new drives of the same size, and new larger drives?) correct?
Yes and yes, but if you add larger drives they'll be used as if they were the same size as the smallest disk in the array: if you have 6 1TB disks in raid 5 and add a new 1.5 TB disk, you'll end up with a 6TB (nominally) array. This is common to all raid controllers I know of.
Also, with the following hardware could I expect at least 60 MB/s over gigabit ethernet with files over a few MB's in size?
Yes, with one possible exception: you may need to buy an Intel gigabit ethernet card on pci express to get there. The number of people I know whose problems have been solved by moving away from Realtek is immense. If you decide to forgo it for the time being and run into problems, test the array locally with bonnie++ or something before running out and buying a card, but it's probably the next step.
 
Yes and yes, but if you add larger drives they'll be used as if they were the same size as the smallest disk in the array: if you have 6 1TB disks in raid 5 and add a new 1.5 TB disk, you'll end up with a 6TB (nominally) array. This is common to all raid controllers I know of.
.

I figured as much, but if I do that 6 times could I get the full size of the new drives? I'd either have to do that or add a number of pci-sata cards to create a new array on in the future to expand

Thanks for the suggestion about the network card; I'll probably get a few of those along with an 8-port HP gigabit switch
 
I've not experimented with hardware raid at all, but the second box in my sig is running that array on linux software raid. I can saturate gigabit with that disk array and even during writes (where raid5/6 are stressed the most) cpu usage says around 10% or so. Doesn't really matter because its a dedicated box. My array started with 2 500 gb drives in raid 1 and grew to 3 in raid 5 then to 4 and then to 8 in raid 6 like you see now. With linux software raid all of the expansions I have performed did not require me to take the array offline with the exception of physically adding the drive and having to resize the partition (the ext2 resizer needs the partition to be unmounted before it can work properly). mdadm handles all of the rebuilding in the background and is transparent to the user. I've not noticed a drop in throughput either while the array is rebuilding.

This is a good howto that helped me to get started with my array: http://linux-raid.osdl.org/index.php/Linux_Raid
 
I have done a similar thing, but I was using raid 1 pairs due to the time/size differences between my drive purchases (2 x 40, 2 x 80, 2 x 160). You can configure the cf card as /boot and use the mount option "noatime" in the /etc/fstab to prevent the OS from updating the access times on files on the cf card and prematurely wearing out the drive.

My last setup I went rounds with trying to get that setup working with a jmicron controller and grub. I eventually gave up and I use RAID 1 on ext2 for /boot with two 80 gig drives (also housing / and swap). I use ext2 so that I don't have to deal with problems from the bootloader.

A few pieces of advise:

1. I don't like ext3 and have had lots of problems with it, you might want to look at one of the other fine journaled file systems. I have good experience with XFS, plus it's well suited to large media. YMMV

2. You may want to create your filesystem directly on the disk without creating a partition. The nice thing about that with an online expansion capable RAID system, you can grow your disk and don't have to worry about finding a partition format that will support multi-terabyte filesystems without limitations and that is one less step in what you are doing. I do this with my hardware controller and have expanded it and grown my primary filesystem without issue.

3. I would also recommend that you create a separate partition for your huge pile of files and one for the root file system of the OS. This is nice for the situation where you want to change linux distros or if you do something like change from 32 bit to 64 bit. You can wipe out your OS partition and know your data is safe on the other partition.

4. If you RAID your swap I would recomend (for a home machine) to use md to create a raid drive to put swap on instead of using two partitions and letting the kernel stripe swap writes across the partitions. If your drive with swap crashes and it's striped by the kernel, the box will probably crash, potentially trashing anything active on the box. If its built on a MD raid, you can just swap the disk without rebooting the system. As dirt ass cheap as small drives are these days, I find it worth the extra disk and power... You could also just put a buttload of RAM in the box and disable swap :)
 
With 4GB of RAM, couldn't he just omit swap altogether if it is going to be a dedicated file server?
 
Back
Top