FreeNAS isn't what I expected?

[jgreco]

"Never go wrong?" Doubtful. However, speaking from experience here it's been an anomaly to RMA a Supermicro board.

Depending on the seller you are slightly more likely to get a board that has not been abused in stupid ways. People typically buy Supermicro to build servers, and that typically works out to be companies. They want their servers to work reliably so they're not likely to be assembling their servers on the kitchen counter.

 

[Whattteva]

I take it supermicro boards are a fave as they are server grade boards, support ECC ram and never go wrong? What would your thoughts be towards buying one second hand?

"Never go wrong" is a very bold claim... I don't think there are any manufacturers that even dares to claim such... If such things were possible, we would not need warranties and redundancy measures.

 

[wintermute000]

The key is driver support/compatibility. The boards (and their components - controllers, NICs, etc.) people have used on this forum are known to play nice with freenas/freebsd.

You go with random board XYZ to find that the SATA controller misbehaves for example, you're out of luck, there's no packaged windows driver to click click click fix the issue.

 

[pschatz100]

The issue, really, is that FreeBSD does not support as wide a range of hardware as does Linux, so you have to do your homework and make good choices. There are non-SuperMicro boards that will do the trick and there are AMD configurations as well that work OK. But don't expect a $30 motherboard to be sufficient. The ZFS file system also imposes some hardware requirements. If you read the stickies about hardware requirements and look at the configurations posted in the signatures of the frequent posters, you'll learn what works and what to avoid.

As to the question of second hand Supermicro boards: Why not? When I decided to learn about servers, I picked up a used Supermicro X9SCL+-F (with a two week warranty) for $89. I paired it with a Celeron that I purchased for $34. 8Gb of ECC RAM cost me about $90 at the time. These costs are small compared to the disks and, after two years, the machine still runs great.

The machine in my signature is my home media server and NAS. I splurged on that one because I wanted to do transcoding of HD material - hence the Core i3 processor. A Pentium processor would have been quite a bit less expensive and quite sufficient for NAS purposes.

The issue about ECC versus nonECC memory is really a function of the way ZFS works. If you read the stickies, you'll know why. I think the people who complain most on this issue probably went out and bought the cheapest possible memory. When I last purchased memory (I am a fan of Crucial) 8Gb ECC memory was priced about $15 higher than the nonECC sticks. To me, this is not a showstopper. The biggest problem was that local stores did not have it in stock, so I had to order it online.

 

[Knowltey]

I take it supermicro boards are a fave as they are server grade boards, support ECC ram and never go wrong? What would your thoughts be towards buying one second hand?

More generally since we know they quite often work with FreeNAS, so the community tendency just continues towards them as a favourite for server boards. Their pricing for features tends to be the best bang for the buck as well from what I've seen.

Buying second hand should be fine as long as you still make sure to do the proper burn in and testing phase on your own as well rather than relying on the previous owner's word that it actually works properly. If it's older too it may be running an older BIOS and may not recognize some newer processors, so try to make sure the BIOS is as recent as you can before you get the board unless you are getting it in combo with the processor that it's been running.

 

[AlainD]

www.freenas.org -> about freenas -> faq

http://www.freenas.org/faq/

Thanks, that's indeed the mentioned FAQ page.

 

[Trif55]

Yea, I'm definitely sold on buying some ECC memory when I build a storage server, as you say, the cost compared to non-ECC is not that much higher and having read the stickies explaining about such matters - specifically the part about hardware RAID controllers always using ECC memory - its really rather obvious we should use it if we want to run RAID using main memory (software raid) :)

I would definitely do burn in testing, that's one of the first things I Googled when FreeNAS booted - "how do I test and burn in my disks?" - I have to say, I think the FreeNAS UI would really benefit from some "best practice" burn in scripts built in, but maybe its best left to the CLI?

"Never go wrong?" Doubtful. However, speaking from experience here it's been an anomaly to RMA a Supermicro board.

Depending on the seller you are slightly more likely to get a board that has not been abused in stupid ways. People typically buy Supermicro to build servers, and that typically works out to be companies. They want their servers to work reliably so they're not likely to be assembling their servers on the kitchen counter.

I suppose I didn't mean "Never" but certainly have a truly server level MTBF :)

On the topic of acceptable failures, reading about ECC ram it says it just halts the system if there's a problem it can't correct, I realise that's better than the alternative - trashing the pool, however would a hard stop still be likely to cause issues/loss of data? (also relevant for choosing quality of PSU/UPS etc - how well would a system deal with a hard stop?)

 

[Ericloewe]

Yea, I'm definitely sold on buying some ECC memory when I build a storage server, as you say, the cost compared to non-ECC is not that much higher and having read the stickies explaining about such matters - specifically the part about hardware RAID controllers always using ECC memory - its really rather obvious we should use it if we want to run RAID using main memory (software raid) :)

I would definitely do burn in testing, that's one of the first things I Googled when FreeNAS booted - "how do I test and burn in my disks?" - I have to say, I think the FreeNAS UI would really benefit from some "best practice" burn in scripts built in, but maybe its best left to the CLI?



I suppose I didn't mean "Never" but certainly have a truly server level MTBF :)

On the topic of acceptable failures, reading about ECC ram it says it just halts the system if there's a problem it can't correct, I realise that's better than the alternative - trashing the pool, however would a hard stop still be likely to cause issues/loss of data? (also relevant for choosing quality of PSU/UPS etc - how well would a system deal with a hard stop?)

Such a problem is quite unlikely. Should it happen, it shouldn't be too different from a kernel panic, in terms of effects.

 

[cyberjock]

On the topic of acceptable failures, reading about ECC ram it says it just halts the system if there's a problem it can't correct, I realise that's better than the alternative - trashing the pool, however would a hard stop still be likely to cause issues/loss of data? (also relevant for choosing quality of PSU/UPS etc - how well would a system deal with a hard stop?)

It handles it in a similar fashion as hitting the reset button, a sudden loss of power, or a kernel panic. While definitely not something you "want" to happen, it's the lesser of the two evils.

 

[solarisguy]

Are correctable ECC errors written to FreeNAS logs? Anyone willing to share their entries?

 

[Ericloewe]

Are correctable ECC errors written to FreeNAS logs? Anyone willing to share their entries?

They're written to the IPMI log, but I don't think the OS is informed.

 

[solarisguy]

They're written to the IPMI log, but I don't think the OS is informed.

So..., no IPMI = no knowledge about ECC correctable errors (CE)... ?

 

[cyberjock]

They are written to the IPMI log but they are also provided to the OS on Supermicro hardware. I can't vouch for anything except Supermicro though. You'll get them as part of FreeNAS' 3am nightly maintenance if there's a problem. :)

Of course, if you're one of those people that still shutdown their server at night you'll never know... so good luck to you.

 

[solarisguy]

So..., is there anybody who has Supermicro and correctable ECC errors from FreeNAS logs to share?

Thank you in advance!

 

[cyberjock]

I don't have one that I can share right now but I've dealt with several people that have gotten the emails and I've seen them myself.

If I'm not mistaken we've had 1 or 2 users in the forum post asking about the emails that said that there were correctable ECC errors and what they should do.

 

[Knowltey]

I don't have one that I can share right now but I've dealt with several people that have gotten the emails and I've seen them myself.

If I'm not mistaken we've had 1 or 2 users in the forum post asking about the emails that said that there were correctable ECC errors and what they should do.

Would you happen to know what the messages look like, keywords, etc so we can set up filters to alert us to those particular e-mails when they come across?

 

[cyberjock]

It looked like a regular email you get when something like a scrub starts or a zpool becomes degraded.

If you filter for ECC or correctable that is what I would do.