Disk management advice needed

[Sn00zE]

Hi guys,
I have build myself a NAS box. Have a HP Microserver with some mods to it. Freenas will be running of USB flash from internal USB.
The Microserver has the standard 4x 3.5" bays in the front, added a raid card to pci-e and a Vantec MRK-425ST EZ-Swap F4 Mobile Rack 4x 2.5" SATA6G in the 5.25" Rom bay, also have a cable running from the esata to internally for 1 more extra drive behind the Rom bay.

I am planning to add the following:
4x 3.5" SATA 2TB drives (3TB if i get them cheap) - in the front bays
4x 2.5" SATA 1TB drives in the Vantec EZ-Swap bays
1x 2.5" SATA 1TB (could be a 2TB also) behind the Vantec EZ-Swap bay

Now my question is how to configure the pool setup or manage the disks the best way? Anyone got some ideas?

 

[jgreco]

What's your goal? How tolerant are you to data loss? Etc.

The general setup suggests a vdev made out of the 3.5" drives and a vdev made out of the 2.5" drives. You could then have both vdevs as a single pool or two separate pools.

How you configure those vdevs is primarily a function of tolerance for data loss. Personally I like RAIDZ2 or RAIDZ3.

 

[Sn00zE]

jgreco thanks for you reply. It sort off confirms what i have been thinking about doing.

Data loss is not such a big issue, its mainly for media streaming. I am looking at adding a backup add-in to Freenas but the data from this will be backed up at least twice weekly on another storage device on the network so no real worries about that..

Was wondering if the 3.5" drives should be combined in a vdev and same goes for the 2.5" drives. Having both vdevs as a single pool is what i would like to do. Does this have fault tolerances for at least 1 drive failure in either vdev then?

Will there be major performance impacts off hand with this configuration? Copying 10GB files at 25mb/s will be a pain! My whole network is Gigabit.

Your own or anyone else willing to share their opinion how you would configure this for one pool/volume is welcome to do so. :)

 

[cyberjock]

You don't seem to have a grasp of this ZFS/RAID/VDEV/fault tolerance thing. Read my guide. It'll explain it ;)

 

[Sn00zE]

You don't seem to have a grasp of this ZFS/RAID/VDEV/fault tolerance thing. Read my guide. It'll explain it ;)

haha thanks noobsauce80, i do understand some basics of freenas by now. Just so much info to grasp. :p more readings, here i go! :D

I'm actually also interested in how you guys that know little more about the system already would configure an overboard HP Micro server like i made! :p

Here she is, remember behind the Vantec EZ-Swap bay is place for one more 2.5" drive

 

[jgreco]

Random thoughts for you to sort through.

Do please make sure you've got 8GB or more in that Microserver. (Yes I know HP says 8GB max, but 16GB works). This is mainly a frustration-prevention step for your benefit.

If you have a workable backup strategy, RAIDZ (which implies RAIDZ1) is perfectly reasonable to consider. Media streaming is not necessarily that intense (it may actually be one of the easier things for a NAS).

RAIDZ2 is generally slower than RAIDZ1. Striped will get you fastest of all, but of course no data redundancy, so any loss means restoring from backups, probably bad. However, RAIDZ1 in your case has one bad side effect: ZFS works best with certain numbers of drives.

RAIDZ1 vdevs should have 3, 5, or 9 drives in each vdev
RAIDZ2 vdevs should have 4, 6, or 10 drives in each vdev
RAIDZ3 vdevs should have 5, 7, or 11 drives in each vdev

You really need 5 devices to get "best" performance out of a RAIDZ1 vdev. Sucks, with all these 4-drive chassis systems, I know. So you could consider making one of your vdevs 5 devices, which you were already considering. That one will be faster than it would have been as a 4 device vdev.

If you combine the two vdevs in a single pool, tolerance for one lost drive in each vdev. Beware: two lost drives in a single vdev can render the whole pool useless. This might be a reason to consider two pools, but if you have backups, probably not.

As far as performance, our N36L is currently suffering abuse as an NFS target for VMware ESXi backups, 4 WD20EARS in RAIDZ2 and appropriate tuning to the task, and it sucks it down at about 400Mbit/s-650Mbit/s.

 

[Sn00zE]

jgreco, you say 16GB ram is working in a HP Microserver?! gota give that a try! Have 8GB now so will get some more soon! Read in a few docs that more ram is better! :D

RAIDZ1 vdevs should have 3, 5, or 9 drives in each vdev
RAIDZ2 vdevs should have 4, 6, or 10 drives in each vdev
RAIDZ3 vdevs should have 5, 7, or 11 drives in each vdev

Agreed it sucks all these chassis only have 4 drive slots. Mmnnn maybe i can squeeze one more 3.5" in the server to have 5x 3.5" drives also.. Will tear it down and see what i can do. Else i will have to sell the whole thing and start over with a new chassis!

Backups will be done of important data stored on the pool at least twice weekly. Loss off media like movie's, series and music will be acceptable. Photo's and other data will always be backed up! :)

 

[Sn00zE]

@noobsauce80 the PowerPoint presentation is awesome! Now lots more sense have been made!! xD Thank you very much! :D

 

[jgreco]

jgreco, you say 16GB ram is working in a HP Microserver?! gota give that a try! Have 8GB now so will get some more soon! Read in a few docs that more ram is better! :D

Code:

Copyright (c) 1992-2012 The FreeBSD Project.
Copyright (c) 1979, 1980, 1983, 1986, 1988, 1989, 1991, 1992, 1993, 1994
        The Regents of the University of California. All rights reserved.
FreeBSD is a registered trademark of The FreeBSD Foundation.
FreeBSD 8.3-RELEASE-p4 #0 r241984M: Wed Oct 24 00:57:10 PDT 2012
    root@build.ixsystems.com:/usr/home/jpaetzel/8.3.0-RELEASE/os-base/amd64/usr/home/jpaetzel/8.3.0-RELEASE/FreeBSD/src/sys/FREENAS.amd64 amd64
Timecounter "i8254" frequency 1193182 Hz quality 0
CPU: AMD Athlon(tm) II Neo N36L Dual-Core Processor (1297.85-MHz K8-class CPU)
  Origin = "AuthenticAMD"  Id = 0x100f63  Family = 10  Model = 6  Stepping = 3
  Features=0x178bfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CLFLUSH,MMX,FXSR,SSE,SSE2,HTT>
  Features2=0x802009<SSE3,MON,CX16,POPCNT>
  AMD Features=0xee500800<SYSCALL,NX,MMX+,FFXSR,Page1GB,RDTSCP,LM,3DNow!+,3DNow!>
  AMD Features2=0x8377f<LAHF,CMP,SVM,ExtAPIC,CR8,ABM,SSE4A,Prefetch,OSVW,IBS,SKINIT,WDT,NodeId>
  TSC: P-state invariant
real memory  = 17179869184 (16384 MB)
avail memory = 16415256576 (15654 MB)
ACPI APIC Table: <HP     ProLiant>
FreeBSD/SMP: Multiprocessor System Detected: 2 CPUs

HP lies. But hey, so does Intel/Supermicro (Atom D525 board, claimed 4GB, does 8GB), Apple (Mac mini 2011, claimed 8GB, does 16GB), etc. It's really gotten to the point where you can't believe what a manufacturer says anymore. If you want memory known to work with your NxxL, use Google for "microserver n36l 16gb" or whatever your model is and see what people say.

More RAM is usually helpful but at 8GB you probably don't need to panic and go racing out to get more. Just do it if and when you feel the thing's a little slower than you'd like.

Most documents will tell you more RAM is better. That's about 95% true. I will mention in passing that there can be too much of a good thing; I spent some significant time chasing down problems in the ZFS txg subsystem resulting from "excessive" (32GB) memory combined with a slow pool.