BUILD Haswell build.

[FREAKJAM]

Very interesting debates here, really worth to read.
I still have a question tho about TLER:

Using TLER is not required, your array will work fine without TLER on FreeBSD+ZFS, using the new AHCI driver. You only absolutely need TLER if your server is a mission-critical production server which cannot afford downtime for even a minute. Drives without TLER may take up to a minute to recover any weak sector (before it becomes bad) - these small hickups are nothing to worry about in home systems and anything that isn't crucial to the survival of humankind.

I'm planning to run a OS supporting ZFS (FreeNAS probably) in a VM (ESXi). I need to passthrough my disks to the VM using the onboard LSI 2308 in my case. I'm reading stories that I must use TLER, since i'm using the hardware raid controller. Is this true? Also reading that LSI is an exception, since i can setup IO timeouts in the LSI 2308 bios:

IO Timeout for [selected] Devices: This field displays the time (in seconds)
which the host uses to timeout IOs for the following device types.
The default setting is 10.
- 00h (Direct Access)
- 01h (Sequential Access)
- 04h (Write Once)
- 07h (Optical)
- 0Eh (Simplified Direct Access)

Any recommendations bout what harddrive is best and wether I should use TLER or not? I've read that TLER should be disabled at all times when using ZFS (when using no hardware raid). After reading the stories in this thread, im going to setup 4x4TB in RAIDZ2.

 

[cyberjock]

TLER prevents a disk from being dropped from an array if a hard disk spends too much time doing read/write retries due to a bad sector. For most hard drives, they will give up and just report the bad sector to the system after 10seconds to 1 minute(depends on brand, model, and firmware revision). If it takes the disk 1 minute but the timeout on the SATA/RAID controller is 30 seconds, your disk can expect to be dropped from the system. That's obviously not cool as the first bad sector your disk has will drop it from the pool. TLER fixes that by not doing an exhaustive attempt to read the sector and after 3-7 seconds just reports the sector as bad and lets your RAID/ZFS handle the error appropriately. Your hardware configuration really determines the "need" for TLER. Generally speaking, I've had good results with non-TLER drives on SATA controllers and TLER supported disks in hardware RAIDs. It can be very hit and miss and controllers that have the timeout value set too low will pretty much be a problem child with any disk that has a bad sector and doesn't have TLER. But controllers that have the timeout value set really high will never have a problem regardless of what hard drive you use.

So the bottom line, do your homework and read up on what actual users of your RAID say.

Keep in mind that running FreeNAS in a VM is not the best solution. The number of users that have lost their pools suddenly without warning is high enough that http://forums.freenas.org/threads/p...duction-as-a-virtual-machine.12484/#post58364 and http://forums.freenas.org/threads/a...ide-to-not-completely-losing-your-data.12714/ were written to warn off would-be doers.

Also keep in mind that using hardware RAID controllers are not recommended with ZFS because they often hide SMART monitoring, SMART testing, and often don't report read/write errors to the OS. For that reason the M1015 is the recommended card for FreeNAS as its an HBA. This is discussed in the Hardware section stickies and the FreeNAS manual.

Most people lately have been going with the WD RED and Seagate NAS version(I forget the exact name) and have had success with them.

 

[jyavenard]

you keep repeating the same thing over and over about people loosing their array due to a URE or lack of ECC RAM.
I'm actually, more and more convinced that this an invention of yours, because certainly I'm yet to see a single case on this forum for this to actually be true...

If ZFS was unable to be mounted following a single bit error (which would have been corrected by the use of ECC RAM) or a URE; that would make it the most unreliable file system ever...

luckily it's not...

Maybe you've read about it when you started playing with ZFS (only last year I believe), and continued to propage this FUD.

As far as what the URE is, in this context (there are obviously several types); unfortunately, you got it wrong. It's unrecoverable at the time it's being read. Read the article you link to in your signature... Remember that probability, one bit in 10^14 ? It's a probability that when you read your disk; it will fail for one in 10^14 bits.

I wonder how they made you a moderator on these forums; it's rather obvious to me that you actually know very little, but are extremely vocal and can copy/paste very well.

 

[FREAKJAM]

thanks cyberjock.

I already stated in my start post that I am aware it is not recommended to run FreeNAS is a VM. You should read my first post too! *wink*
Im reading successful stories bout the new supermicro board (X10SL7) combined with the LSI 2308, thus the reason for not using the M1015.

Reading really good stories bout the new WD greens. They don't come with TLER, but very low power consumption and really quiet. Other disadvantage is that is only comes with 2 years of guarantee.

PS
I'v been reading forums for almost 6 weeks now, winning knowledge and advice, before i will actually buy my setup.
So yeah, i'm doing my homework =) Only need a few teachers here and there to educate me! =D

 

[cyberjock]

I can trash NTFS in a single bit too. In fact, I can trash every file system out there in a single bit. It's actually quite easy to do. The fact that you don't know how shows the lack of knowledge. I'll leave you to study up on file systems and whatnot and figure out how to do it yourself. ;)

They made me a moderator because I have a clue Mr. 25 whole messages in 12 days... so clearly an authority we should all believe without question.

Anyway, I'm done here. There is no intelligence to be found in this thread. I am glad that when people read your posts versus mine they'll hopefully consider a moderators comment before a "new member". Might save a few people from losing their data by listening to your opinions with no facts.

 

[jyavenard]

TLER prevents a disk from being dropped from an array if a hard disk spends too much time doing read/write retries due to a bad sector.

That's all good for hardware RAID cards.

I can't think of a single software RAID solution for which it actually matters (be it linux or BSD, simply because how their controller drivers are behaving).

TLER allows the disk to still respond to the controller that it is still live, just busy accessing a sector.

A RED drive is the same hardware as a WD Green drive, with a tweaked firmware implementing TLER and that won't park heads when idle as quickly (that can be changed with a green drive too). It also comes with an extended warranty.
Same deal between the Seagate Desktop and the Seagate "NAS": same hard drive, different firmware.

TLER default varies between manufacturers; WD and Seagate have different default settings.

Personally, I think you should focus on the warranty, because in your typical home environment, it won't make a difference.
Seagate NAS disk is very new, WD Red has been going for more than a year now: not sure if that matters to you.

 

[FREAKJAM]

Anyway, I'm done here. There is no intelligence to be found in this thread

I'm really hoping that message isn't towards me. I registered on this forum cause of the expertise i'v read in your and jgreco his posts mostly. Like i said i'm totally new to ZFS.

As stated, i'm totally new with ZFS & FreeNAS, so correct me when wrong if/when i'm making wrong statements.

My start post is very detailed and nobody actually answered most of my questions and people kinda hijacked the thread, which ended in a heavy discussion.
I did not plan for this to happen and I was asking for advice only.

 

[FREAKJAM]

For the people who still are interested:

I returned my case to the manufacturer, and went with a bigger case, the Bitfenix Phenom. It can hold up to 6 drives and 1 SSD, so it's a perfect fit. Currently waiting for the CPU and RAM to arrive, after that it's time to start building :)

Placed the 120GB SSD in my current HTPC, so i can re-use the other drives (2x2TB). I want to make VM/ZFS snapshots and save them to my zfspool and periodically copy the VM snapshots from ZFS pool to my HTPC so they are backupped on a different system.

Going for a RAIDZ2 (6x3TB) setup now:
1Intel Xeon E3-1230 V3 Boxed
1Supermicro X10SL7-F
1Kingston ValueRAM KVR16E11K4/32 (32GB)
6WD Red SATA 6 Gb/s WD30EFRX, 3TB
1BitFenix Shadow
7Sharkoon SATA III standard (50cm)
1Seasonic G-Serie 360Watt
2BitFenix Adapter (Molex to 4x SATA)
1Crucial 2.5" M500 240GB (ESXi datastore)
1USB-stick (ESXi boot)

Going

 

[MrMadMan]

Any updates on the build? I'm doing something very similar (check my thread if interested) and I'd love to hear your experiences :)

 

[FREAKJAM]

Very sadly my motherboard had damaged/broken pins on the cpu socket and I did not notice the damage untill I was just about to place the cpu in the socket.
Having a hard time convincing the reseller that I think it's impossible that i might have damaged the pins after I removed the socket protector.

their statement is damaged pins == no warranty :(

 

[MrMadMan]

Damn. That sucks :(

But that's usually how the story goes when it comes to broken pins...

 

[cyberjock]

Yeah, every single manufacturer says damaged socket = warranty void. That's precisely why it comes with the socket protector. Most manufacturers won't even warranty your board if you send it back to them via an RMA for another reason without the socket protector installed.

 

[jgreco]

So that's when you very carefully try to fix the pins and then RMA it as a dead board. Heh.

 

[FREAKJAM]

Supermicro is sending me a new board after all or either going to repair the cpu socket. Law in Holland states that the reseller must prove up to 6 months after a purchase was made, that the product was delivered without any damage. I found it very hard to believe that I damaged those pins. I removed the protector really carefully. I was simply looking away to my precious xeon cpu right after I folded over the protector, so yeah, I did not see the damage right away.

My reseller refused to cooperate at first, but after mailing a bit with their supplier and with supermicro, they decided to send me a new one or repair it after all.