Stuck before I start - Downloading 9.2.1.8 USB Image

[cyberjock]

I wouldn't expect to find server grade system for the ISS or spacecrafts and satellites going into space. Definitely server grade is not going to perform any better than desktop computers. You will need harden hardware for that.

Actually, you'd be pretty amazed at the stuff that goes in space. It always, always, always has ECC (or some other correction type) because of radiation in space. It's almost always custom built for a specific purpose, making it outrageously expensive. A few (mostly cheaper one-time short-lived) satellites use hardware like Supermicro and encase it in shielding with a custom BIOS that does things like crazy memory scrubbing, low power options that are a bit more aggressive, etc. Generally, they take a generic design for what you and I call " server-grade" motherboard (or whatever part you are talking about) and they have the manufacturer setup custom firmware that forces certain settings, disables other ones, etc. What's funny is that they are *way* behind the times when it comes to using the latest tech. In fact, one of the biggest problems for space has been that transistor size has shrunk (duh) which means that a single radiation event causes a significant percentage of a transistor to be ionized (sometimes resulting in a "fire" of a transistor that is supposed to be off). This literally translates into newer CPUs (with smaller transistors) to have much higher error rates than older (and obviously slower) CPUs and in some cases are not reliable enough to even be used. So how do you counter it? Add more shielding. Oh, but shielding adds weight which quickly adds cost to the satellite to get into space. So you have this very delicate balance of CPU processing power and shielding to figure out how to make everything work. Not everyone can take something with the equivalent processing power of an E3-1230v3 and throw it into space while avoiding a pricetag that could exceed the GDP of some countries. ;)

I remember reading a few years ago that Intel had finally pulled the plug on manufacturing the 486 and Pentium lines for hostile environments (they had a specific name for it, but I forget what the word was) like space and such. These weren't your typical 486 and Pentiums though. I want to say this was 2011 or so when I read about it. So just imagine the fact that satellites of 2011 were still being made with those and think about that! LOL

One of the problem with technology in space is that as technology here on earth has progressed it has really hurt the space tech industry because they can't grab off the shelf parts like they did in 1990 and then use them as test platforms. You're doing more and more all-custom designs and demanding transistors of a size from a decade ago so you can minimize shielding for cost reasons.

If you ever get a chance to talk to a NASA engineer that has had to design some of that or got a tour of the stuff the public doesn't get to see, definitely do! I had a rare opportunity due to my military experience and it was something I will never forget.

 

[Apollo]

You do get high end electronic components such as space grade FPGA's, even lithium batteries in space though.

 

[solarisguy]

@Apollo, you need to read about ZFS and ECC RAM. This is not an opinion or an attitude. And it has nothing to do with FreeNAS. It is relevant for both ZFS and OpenZFS.

 

[solarisguy]

@Palmy, it turned out to be a good practice to zero the USB memory device before a new FreeNAS image is written to it. At least zero everything beyond first 2 * 942MB = 1884 MB

 

[Tywin]

They use standard ThinkPads on the ISS (not for e.g. life support and comms, but as workstations and terminals). I think they switched to something else recently, though I can't remember what..

Actually, you'd be pretty amazed at the stuff that goes in space. It always, always, always has ECC (or some other correction type) because of radiation in space.

Sure, often, but not always. It entirely depends on the application.

It's almost always custom built for a specific purpose, making it outrageously expensive. A few (mostly cheaper one-time short-lived) satellites use hardware like Supermicro and encase it in shielding with a custom BIOS that does things like crazy memory scrubbing, low power options that are a bit more aggressive, etc. Generally, they take a generic design for what you and I call " server-grade" motherboard (or whatever part you are talking about) and they have the manufacturer setup custom firmware that forces certain settings, disables other ones, etc.

Actually, our recent crop of payload computers are simple off-the-shelf single-board computers. Yes to ECC memory, but otherwise, basically stock. As for shielding, they get a few mm of Al from the box they are in, nothing special (no practical amount of shielding will help against a cosmic ray hit). Somewhat amusingly, we have to remove the board's battery cell (it would explode, and also wouldn't work in vacuum); this causes the BOIS to belch, so we had to implement a tiny widget that emulates a USB keyboard, effectively a drinking-bird that hits ENTER on startup to get past the POST warning :p

What's funny is that they are *way* behind the times when it comes to using the latest tech. In fact, one of the biggest problems for space has been that transistor size has shrunk (duh) which means that a single radiation event causes a significant percentage of a transistor to be ionized (sometimes resulting in a "fire" of a transistor that is supposed to be off). This literally translates into newer CPUs (with smaller transistors) to have much higher error rates than older (and obviously slower) CPUs and in some cases are not reliable enough to even be used.

A better way is to radiation-test your hardware before-hand, to select stuff that is inherently resistant to radiation effects (both single event upsets and single event latchups). It turns out you can't draw a direct comparison between feature size and susceptibility, it is heavily process-dependent. Here is some reading, if you are interested. In particular, check out the paragraph "Operating voltage and Latchup" on pg. 6: "Many modern digital components have core voltages around or below 0.9 V, and these can be expected to be latchup free."

So how do you counter it? Add more shielding. Oh, but shielding adds weight which quickly adds cost to the satellite to get into space. So you have this very delicate balance of CPU processing power and shielding to figure out how to make everything work. Not everyone can take something with the equivalent processing power of an E3-1230v3 and throw it into space while avoiding a pricetag that could exceed the GDP of some countries. ;).

Again, depends on your application. Below 100 kg, launch cost is driven more by planform area than by mass. Also, beyond a modicum of shielding (several mm of Al at the outside for LEO), you rapidly hit diminishing returns -- your driving factor becomes high-energy protons, which no practical amount of shielding can protect against. You are better off carefully selecting your components and putting some thought into the overall design.

Edit: Agreed, I don't know of anyone that has flown an E3-1230v3 or equivalent, but it's not far off..

 

[Palmy]

Use SATA drive if you can this will allow to test on a different approach and take USB out of the loop.
I noticed you have DVI and VGA, is it possible the system is switching between either video output?
Are you sure your monitor is not turning off on it's own (maybe not server grade moitor, then I have never seen or heard of server grade monitor as they usually use serial com instead, or IPMI and such with a dedicated workstation, or server station).
Could it be the video resolution of you monitor?
Last, would it be posible to know as much as what is being displayed at POST, this might give us some clue to what is happening. If you could actually take a server grade picture, that would be even better.

Bingo... well isn't this embarrassing... It was flicking over to DVI. I've got a monitor with both inputs so have been flicking from my laptop to the NAS, after reading that I thought I might just swap them and check... sure enough, it's working. Thank you and I shall now have a server grade coffee to celebrate... :)

 

[Apollo]

@Apollo, you need to read about ZFS and ECC RAM. This is not an opinion or an attitude. And it has nothing to do with FreeNAS. It is relevant for both ZFS and OpenZFS.

I know enough about the benefit and pitfall of using ZFS and ECC.
However the attitude is mostly about people denigrating one thing versus another based on their preference and that goes against the forum etiquette, I think.
I welcome criticism and try to respect people's opinion but do not necessarily share their idea or belief.
I just don't like North American coffee.

 

[Apollo]

Bingo... well isn't this embarrassing... It was flicking over to DVI. I've got a monitor with both inputs so have been flicking from my laptop to the NAS, after reading that I thought I might just swap them and check... sure enough, it's working. Thank you and I shall now have a server grade coffee to celebrate... :)

Well you have to think outside of the server grade box and explore all possibilities before pointing the finger on anything. That's what a good engineer is all about.
The message I am trying to convey is to always keep an open mind. Except for North American Coffee. Sorry.

 

[cyberjock]

Well you have to think outside of the server grade box and explore all possibilities before pointing the finger on anything. That's what a good engineer is all about.
The message I am trying to convey is to always keep an open mind. Except for North American Coffee. Sorry.

Yep. Except statistically he'll be back here shortly with some new obscure problem that has zero results when you do a search. That's why we tell people to not look at desktop hardware.

But good luck. Stick it to the man and show us you'll be the one exception. ;)

I personally don't trust my data to the statistical likelihood that the server won't flop and eat my data randomly at some point in the future.

 

[Palmy]

Well you have to think outside of the server grade box and explore all possibilities before pointing the finger on anything. That's what a good engineer is all about.
The message I am trying to convey is to always keep an open mind. Except for North American Coffee. Sorry.

Yep completely agree, and that's what forums are all about. :)

I also agree with the coffee comment, or coffee in any region that starts with Star...

 

[solarisguy]

I just don't like North American coffee.

Funny :)

I think, you meant that you do not like buying/drinking choices of most USA and Canadian coffee consumers :) . There are no coffee plantations in those two countries, but there is quite good Arabica grown in North America... ;)

See :) One short sentence. Two interpretations :)

 

[Apollo]

Most North American Coffee are processed through the use of an Alkaly process, basically ( pun intended ) Chlorine or Bleach to speed up the torefaction process, and the fact that you can see through to the bottom of the cup even with several inch of coffee.
Nothing better, than European coffee such as French, Italian and Turkish. But then the French are great in so many ways, as well as Italian and a large portion of the European community.

 

[Apollo]

Funny :)

I think, you meant that you do not like buying/drinking choices of most USA and Canadian coffee consumers :) . There are no coffee plantations in those two countries, but there is quite good Arabica grown in North America... ;)

See :) One short sentence. Two interpretations :)

Also the fact they add artificial chemicals to create flavors and smell. Like French Vanilla...

 

[solarisguy]

@Apollo, please be more precise... :) North American coffee can have multiple meanings, but a solution to your problem is trivial: any major North American city would have German or Italian or Polish grocery stores with packaged coffee imported from Europe.

With ECC RAM and ZFS, since you know why, you should try to avoid ambiguous messages and be precise, and say something like ECC RAM is not required for a demo/test ZFS installation or ECC RAM does not equal server grade hardware etc. You had presented ECC RAM as an option/choice instead...

P.S.
Cyberjock knows that I am typing this on a desktop with ECC RAM, but he also knows that for quite some time there are no mass produced desktop system with ECC RAM...

 

[cyberjock]

Also the fact they add artificial chemicals to create flavors and smell. Like French Vanilla...

Hey! Let's not talk about hazelnut coffee. Them thar fightin' words! ;)

 

[Apollo]

@Apollo, please be more precise... :) North American coffee can have multiple meanings, but a solution to your problem is trivial: any major North American city would have German or Italian or Polish grocery stores with packaged coffee imported from Europe.

True, but it is also a fact or maybe a believe that European goods are sometime manufactured to the standard of the country it is being shipped to. You will never see French wine with sulfite when you buy it in France. Same with cheese, try to find a true non pasteurized cheese, and not necesarely French. Lots of Italian and Switch chees, Greek and Netherland...
Same with cars like BMW, most of tehem are manufactured in Mexico for the North American market.
Sorry, I have never had Polish coffe.

why[/I], you should try to avoid ambiguous messages and be precise, and say something like ECC RAM is not required for a demo/test ZFS installation or ECC RAM does not equal server grade hardware etc. You had presented ECC RAM as an option/choice instead...

Yes, possibly. I might add that some of the ZFS pitfall, mostly for Desktop non server grade system lies in the idea that ZFS trust the hardware blindly as it is designed to coexist with hardware that is self correcting, ie ECC.
To be honest, and there was a comment on this forum a while back about it I believe, is that ZFS should be more fault tolerant by cross verifying potential errors.
During scrub or simply accessing data from RAM or dataset, ZFS always checks for data integrity, calculating and comparing the sector (block) stored checksum. However, in the case of non ECC hardware, ZFS will treat a single checksum error as a faulty sector and as Cyberjock may have mentionne so many times will cause ZFS to correct the supposedly bad sector with corrupted data. In this instance, ZFS should be made more reliable giving it the ability to second guess the hardware and the data and use a different scheme to validate the data again.
ECC RAM is believed to be more impervious against certain external interference such as solar flare, but these do not occur that often and when they hit, it would be fairly randomized as the energy exiting the electron layer is mostly what causing the change in electronic states that will appear as a 0 or a 1 will not, or should not affect the same molecule in the hardware (not talking about permanent memory damage). Bit walking or shadowing is more prone to occur. Reading the data back from the drive and placing it in a different memory location could help alleviate non ECC memory issues. My point is that the desktop, non grade hardware will be susceptible to hardware errors because ZFS is not designed in such a way.
NIC and network related hardware are not server grade when it come to the internet, maybe they are but they are not in a control environment. However the hardware and firmware, through the different ISO layers takes care of validating the data with checksum. But then it is easy to request new data again, which is more difficult when it come from a drive.

Technically, there is no need to run memory test on ECC RAM. Nonetheless, people feel better doing so. If memory stress program create errors in hardware, it is in my opinion mostly due by variance in product specification and bad hardware design or too tight tolerances. This is a reason why ECC RAM is slower than "anthousiast" RAM, because ECC RAM manufacturer don't want or can't produce server grade hardware because the envelope is too tight to guaranty.

And yet you pay more for ECC because of that extra single bit or multiple bit error correction capability.

Cyberjock[/I] knows that I am typing this on a desktop with ECC RAM, but he also knows that for quite some time there are no mass produced desktop system with ECC RAM...

And yet, back to the rant game, there are no server grade keyboard, otherwise they would have ECC embedded in it. But oh wait, I may not be that correct as my Browser has built in online word checking, but it is not hardware based so not server grade.

 

[solarisguy]

@Apollo, ZFS could have been different, may be it should have been different. However, I am not a ZFS developer (I do not think there are any around here) and our discussions in forums.freenas.org would not change ZFS behaviour in any way. So the truth is that certain non-ECC RAM errors during a scrub can erase one's data. One can easily avoid that by using ECC RAM. Saying otherwise in this forum only confuses those who have limited IT experience.

 

[solarisguy]

Is there a way to delete one's duplicate post?

 

[Apollo]

Is there a way to delete one's duplicate post?

I do not disagree with your remark, and this is one more reason it is imperative to have a backup, as Cyberjock recommended, so that people with no ECC- RAM can still sleep tight.
Freenas is becoming more needy in term of hardware requirement as opposed to it's former days. But people interested in using Freenas may not necesarely be able to buy proper hardware either when thay start. As they learn and gain experience then maybe they will be willing to level up.

 

[Palmy]

I do not disagree with your remark, and this is one more reason it is imperative to have a backup, as Cyberjock recommended, so that people with no ECC- RAM can still sleep tight.
Freenas is becoming more needy in term of hardware requirement as opposed to it's former days. But people interested in using Freenas may not necesarely be able to buy proper hardware either when thay start. As they learn and gain experience then maybe they will be willing to level up.

It's not even a case of being "able to buy", but the choice of freenas vs Synology/QNAP is where I ended up. If I was going to spend a fortune on hardware, I'd just buy a Synology NAS and know I have hardware/software support. I didn't want to spend that, and the FAQ's showed me (as well as multiple Videos) that desktop grade parts were perfectly adequate. Most people won't spend ours on a forum until they have a problem.

But as of now, I have freenas running, a disk array with a CIFS share published and good non-american bleached coffee... We'll see how long before my poor choice in hardware causes issues. :)

Thanks for the feedback and help. :D