General > General Technical Chat

eBay's purchase protection program doesn't really work!

<< < (12/14) > >>

DiTBho:

--- Quote from: vad on September 27, 2023, 01:38:28 am ---Instead of using eight 500GB drives, you can opt for two 4TB WD Red Plus CMR drives, which would cost about 20 Euros per terabyte. These drives are brand new and come with a manufacturer's warranty.

If you plan to run the drives 24/7 in your NAS, you should also consider power consumption. HGST drives are power-hungry, idling at 5W or higher, whereas the WD40EFPX idles at 3.1W.

Eight HGST drives would consume 350 kWh per year, while two Red Plus drives would consume 54 kWh per year. I'm not sure about the efficiency of your NAS's PSU or your electricity costs, but saving 300-400 kWh or so per year would reduce your carbon footprint.

--- End quote ---

The NAS that I built in 2010 is a bit particular, it's made up of 3 nodes, each based on PPC40x, managing each a pair of disks in soft-RAID-1 (2xnode=2x3=6 disks), provided by a SiliconImage PCI chip. The nodes are controlled by a little SBC that is always online, consumes less than 5Watt, monitors the network traffic entering the NAS, if it notices a request for one of the three nodes that is offline (off), forces a softboot to the node, and in less than 12 seconds (bootstrap time) it satisfies the request. Similarly, if it sees no activity on a node within 40 minutes, it requests shutdown and turns off both the node and the hard disks connected to it.

Periodically (during night) also forces the bootstrap of the mediatower, and also organizes backups according to different levels. From daily backups, on DDS tape, to weekly backups on CDRW, to monthly backups on DVDRAM


--- Code: ---.          _______
          |       |
network === ctrl0 |                     _______
          |       |                    |A      |--- sata.ch0----- hdd0
          |       === lan0 & uart0 === | node0 |--- sata.ch1----- hdd1
          |       |                    |_______|--- pata_ramdisk
          |       |                    |B      |--- sata.ch0----- hdd2
          |       === lan1 & uart1 === | node1 |--- sata.ch1----- hdd3
          |       |                    |_______|--- pata_ramdisk
          |       |                    |C      |--- sata.ch0----- hdd4
          |       === lan2 & uart2 === | node2 |--- sata.ch1----- hdd5
          |       |                    |_______|--- pata_ramdisk
          |       |                     _______________
          |       |                    |D              |
          |       === lan3 =========== |  media_tower  |
          |       |                    |   SCSI        |
          |       |                    |   + ramdisk   |
          |       |                    |   + media     |
          |       |                    |     + DVD-RAM |
          |       |                    |     + CDRW650 |
          |       |                    |     + DDS-4   |
          |       |                    |_______________|
          |_______|
             |
             | uart4
          ___|__________
         |              |
         |  power unit  |
         |              |
         |              |
         |      12V,5V ---- group.A { node0, hdd0, hdd1 }
         |      12V,5V ---- group.B { node1, hdd2, hdd3 }
         |      12V,5V ---- group.C { node2, hdd4, hdd5 }
         |      5V_TTL ---- group.D { media_tower } (ATX poweron)
         |______________|

--- End code ---


--- Code: ---.       ___________
       | bay0      |
       |     ______|_
       |    |  hdd0  |------ node0.sata.ch0
       |    |A_______|
       |    |  hdd1  |------ node0.sata.ch1
       |    |A_______|
       |    |  hdd2  |------ node1.sata.ch0
       |    |B_______|
       |___________
       | bay1      |
       |     ______|_
       |    |  hdd3  |------ node1.sata.ch1
       |    |B_______|
       |    |  hdd4  |------ node2.sata.ch0
       |    |C_______|
       |    |  hdd5  |------ node2.sata.ch1
       |    |C_______|
       |___________|

--- End code ---

Through an internal 1000Mbps Ethernet line managed by a pair of Tulip chips (with dedicated network processors and queues, but limited to 20Mbyte/sec) the data flows from the disks of one of the three nodes to the SCSI ramdisk (8GB) of the mediatower, only then does the backup begin.

That NAS basically manages source archives in C and assembly, as well as vhdl, latek and txt files, vector files and some pictures, all things that never exceed 10Mbytes per file. I have developed software to track updates, so ONLY what ( = a project) has changed is backed up.

A 4GB DVDRAM fits perfectly into my monthly archives.

Made with late 90s and early 2000s tecnology, it had a troubled and controversial start, but this system has been working perfectly for 5 years now.

Why 3 sets of disks? Because things are compartmentalized into 3 big categories, which must NEVER be online at the same time, this is due to my paranoia.

vad:
Wow, that's impressive. Good luck with your project!

DiTBho:



--- Quote ---I wouldn't buy used HDDs on eBay, but in this situation, there's no technical reason to keep drives with zero hours and return used ones.

--- End quote ---

I made a second call to customer service, they replied something similar again.
So if the category is not "new", but "opened never used", it is not a problem for them if the disk has service hours unless the disk is actually "DOA".

The girl on the phone said exactly that, asking if one or more HDDs were "DOA?", "Dead On Arrival?"

as a corollary I conclude that

* on eBay, "opened never used" can be returned only if DOA, next time you have to say that in the motivation, item not as description because not working
* It's better to buy on Amazon, selective purchases, you need 8 disks? you make eight different orders, so you can selective return defecitve parts, otherwise you have to return everything; if something goes wrong Amazon doesn't make any question, you return the disc, and they'll refund you. Point
* It is better to pay directly sellers with Paypal, without going through eBay, which in the end only adds more costs and hassles
* the best solution is to find a serious dealer, like the mentioned Pitbull-something for SCSI parts, and pay him/her with Paypal.

DiTBho:
except the last 5 years when I use the NAS regularly, the 13-5 years before that my HDDs were used periodically but not continuously over the time. Say, 4 days nonstop, then unused for a couple of months, then again 1 week continuously and unused for another couple of months, and so on.

So, you know at least that they are not "dead on arrival".


--- Quote from: vad on September 27, 2023, 12:56:31 am ---The failure rate curve of hard drives typically exhibits a bathtub shape. All other things being equal (no bad blocks, no SMART errors), drives that have been sitting on a shelf unused for 13 years are likely less reliable than those with several thousand hours of use. This is due to the risks associated with child mortality, in my opinion.

--- End quote ---

What would be the problem with an HDD that remains unused for 13 years?
You know, you tested it for a couple of days after purchasing it, and it showed no problems.

The mechanical parts? like the engine bearings?
The electric parts? are the contacts between the driver and the motor and/or sensors oxidising?

In road bicycle hubs the problem is the grease between cones and spheres which "cements" (it hardens). All you have to do is dismantle everything, degrease, put the grease back in, and the hub runs like new again.

Unfortunately you cannot do the same with an HDD, because it must be opened in a dust-free environment

DiTBho:
Found this article:


--- Quote ---Hard Disks Don’t Die Young Anymore - Aug 2, 2022

The old hard disk drive is disintegrating in space. Modern Hard disks are more reliable than their reputation suggests

“Infant mortality” in hard disks has become a thing of the past, according to online backup firm Backblaze.

Backblaze has more than 200,000 hard disks being used to store customers’ data and regularly reports on the reliability of individual models. The latest report shows that traditional spinning hard disks are nowhere near as fragile as their reputation suggests.

Only three of the 27 models of hard disk reported on by Backblaze had an annualized failure rate of over 2.5%. But all three of those models had a low sample size, meaning Backblaze couldn’t be more than 95% confident of the failure rate.

The overall annualized failure rate of all the disks in Backblaze’s study was only 1.46%. Although that’s up slightly from previous years, the company says this is due to the increased age of its fleet, not diminishing reliability. In short: hard disks very rarely fail, especially in the early years.

Eradicating bad drives
How do the disk manufacturers achieve such low failure rates, with many of the disks in Blackblaze’s fleet now being in active service for five or six years? Backblaze’s Principal Storage Cloud Evangelist, Andy Klein, puts it down to improvements in manufacturing.

“It used to be the drives, just like everything else mechanical and electromechanical followed a bathtub curve”, he said. The bathtub curve sees relatively high “infant mortality” in the early life of a disk, before settling down, only for failure rates to climb again as the disks age.

“What we've seen over the last several years now is the front end of that curve is almost flattened out”, said Klein. “And you see the backside start to push a little bit. So where the failures started to rise after two and a half, three years, now the rate doesn't start to rise until you get to about four years.”

Forbes Daily: Get our best stories, exclusive reporting and essential analysis of the day’s news in your inbox every weekday.

By signing up, you accept and agree to our Terms of Service (including the class action waiver and arbitration provisions), and you acknowledge our Privacy Statement.
Klein believes the early failure rates have been largely eradicated due to better testing of new models by the hard disk manufacturers, who are putting new drives through more intensive tests before putting them on sale.

“Getting a DOA [dead on arrival] drive happens a lot less”, said Klein. “I can’t recall our guys talking about that at all in the last year or two.”

Using AI to predict failures
With drive failures becoming increasingly rare, Backblaze has been experimenting with AI models to predict when drives in its fleet may die.

Although such AI models have had some success, the problem is they don’t easily translate from one model of hard disk to the next. “Every new model creates a different set of characteristics that you have to go and track down”, said Klein.

However, one model created by a third party did identify the characteristics that will show when a drive is likely to carry on working normally for the foreseeable future.

“To have that kind of knowledge is kind of interesting,” said Klein. He said that if the company can deploy a drive and know that “under this set of circumstances, it's got a 99% chance it's going to survive a year, that's a good thing, I'm feeling really good about that.”

That might allow the company to delay maintenance on a particular server, or not replace a disk that would normally be expected to be reaching the end of its lifecycle.

--- End quote ---
(from here)

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod