Popular Post mansr Posted March 11, 2019 Popular Post Share Posted March 11, 2019 25 minutes ago, Shadders said: Not wishing to dampens things - but i have had multiple WD drives fail All drives will fail eventually. All drive manufacturers have had bad models or batches. To improve your chances, pick a model intended for the usage you expect. For an always on system, choose "enterprise" models. If the drives will be spun down or powered off frequently, go for something else. Always keep backups of valuable data. RAID is no substitute for backups, but it does help with availability. tmtomh, sandyk, jtwrace and 1 other 4 Link to comment
Popular Post mansr Posted March 11, 2019 Popular Post Share Posted March 11, 2019 12 minutes ago, Shadders said: Check the following 1-star reviews : https://www.amazon.co.uk/Seagate-inch-Barracuda-Desktop-Drive/dp/B0073Q7GU6 And : https://www.extremetech.com/extreme/222267-seagate-faces-lawsuit-over-3tb-hard-drive-failure-rates There is an issue. Yes, that 3TB model is infamously unreliable. As was the IBM DeathStar back in the day. And many others over the years. Don't buy drive models with a bad reputation, but there's no reason to avoid any of the major manufacturers because of one incident. If you want reliability, buy the slightly more expensive server/enterprise drives. These have much better specs than the consumer models and are meant to run continuously. I've had four 4TB Seagate Constellation ES.3 drives running for nearly 5 years. About two years ago, one of them reported a few bad sectors, so I replaced it as a precaution. You guys are overthinking this. Disk drives are cheap. Treat them as consumables with a half-life of a few years. Use RAID for zero-downtime replacements and backups to cover catastrophic failures. plissken, Jud and tmtomh 2 1 Link to comment
mansr Posted March 11, 2019 Share Posted March 11, 2019 27 minutes ago, ipeverywhere said: Software RAID is bad please try to avoid. What gave you that idea? "Hardware" RAID controllers still have a CPU that runs software. If you're thinking of OS level RAID as opposed to a separate controller, the former is actually preferable. Why? Because the on-disk format is documented. If a "hardware" controller breaks, chances are that model is no longer being made, and the new ones use a different format. All your data is lost, even if the drives are intact. With "software" RAID, you are no longer reliant on a single vendor. Most, if not all, of those little NAS boxes use Linux software RAID, by the way. Link to comment
mansr Posted March 11, 2019 Share Posted March 11, 2019 10 minutes ago, ipeverywhere said: 1, and primarily,) The extra tax on the CPU to maintain/manage the array. My opinion is that the machine processing the audio should be processing the audio and not trying to do much else. I agree that most of the cheaper "RAID" solutions out there are simply using soft RAID under Linux to manage the array. The point here is that it's external and doesn't impact the machine serving the music. The overhead on reading is low (zero for RAID1). At audio data rates, it is utterly negligible. 14 minutes ago, ipeverywhere said: 2) It's less reliable (maybe unreliable in certain situations?). Without diving deep into what a memory backed hardware raid array does compared to a software raid I truly believe the software arrays are significantly less reliable than the hardware option. You're talking about RAID controllers with battery backed RAM. The advantage of those is mostly about performance. Thanks to the battery, writes can be considered complete as soon as the controller has received the data rather than when it has been written to non-volatile storage. For certain workloads, this can make a big difference. Audio streaming is not one of them. The battery also means that on sudden power loss, the very last writes will be preserved whereas without it a handful of blocks may be lost. That difference is of little consequence in most cases since the power failure might as well have happened moments earlier. Either way, the filesystem and applications need to cope with partially written data. Link to comment
mansr Posted March 12, 2019 Share Posted March 12, 2019 3 minutes ago, rando said: For long term unpowered storage. Where an indisputable advantage over spinning drives exists. No slow decline of unacessed data or parked heads means full data recovery after 10 years gathering dust in a bank vault. You should do some reading on data retention in NAND flash. It's not as stable as you think. sandyk 1 Link to comment
mansr Posted March 12, 2019 Share Posted March 12, 2019 2 minutes ago, rando said: @mansr I think both of us agree spinning hard drives 2/3 full have greater chances of retaining data than 5.5/6 full drives. I'm not sure what you mean. Are you saying that drives are more reliable used only to partial capacity? 2 minutes ago, rando said: We may be split down the middle on documentation and practical real world usage of SSD for infrequent backups. Everything I've read about brief to months long retention ability has been positive though. Good controllers do help alleviate the weak points of NAND flash. Months yes, years no. High-density NAND flash relies on the controller periodically reading each page and rewriting it if too many (correctable) errors have accumulated. Briefly powering it up and copying a few files might not be enough either. Link to comment
mansr Posted March 12, 2019 Share Posted March 12, 2019 7 minutes ago, rando said: I am indeed positing HDD are impacted negatively after reaching very near to full capacity. Especially where rarely to never again accessed data is concerned. I can see no reason why this would be the case, nor have I ever heard such a claim before. However, sequential read/write speed is higher at the start of the disk than the end. For that reason, disks are sometimes utilised only to half or ⅔ capacity. Reliability has nothing to do with it. 11 minutes ago, rando said: SSD that had all memory written over a few times beforehand are statistically more robust. This I can accept with some caveats. All NAND flash devices have some bad blocks, so an initial scan like this could reveal a few that were missed in the factory. I doubt it makes much if any difference in practice, though. Bad blocks can develop at any time, and the controller has to deal with them without losing data. For this reason, and (more so) to facilitate wear levelling, an SSD has a considerable amount of extra capacity beyond what is user visible. Even so, if only part of the available capacity is used, the wear levelling could potentially work more efficiently given the extra space. A partially used SSD might thus be more reliable than a full one. Maybe. Link to comment
Popular Post mansr Posted March 24, 2019 Popular Post Share Posted March 24, 2019 2 hours ago, sphinxsix said: I can't blame the drive for the failure if it was connected by mistake to the wrong power supply. I would. Adding protection against a small over-voltage (guessing the proper supply is 12 V and the laptop supply 19 V) would have cost pennies. 2 hours ago, sphinxsix said: I just wonder: 1. Will it be possible to fix it? 2. Will I get my data back? 3. How much will it cost.? Short answers: 1. Maybe. 2. Maybe. 3. Depends on what is fried and on your (or your friends') soldering skills. If you're lucky, only the primary voltage regulator has been fried and the drive itself is intact. In this case, you can simply move the drive to another enclosure. sphinxsix and rando 1 1 Link to comment
mansr Posted March 27, 2019 Share Posted March 27, 2019 9 hours ago, sphinxsix said: I also must say I just love the idea of estimating the scale of a damage by smelling the device The smell of burnt electronics is unmistakable. It's less reliable these days, though, since sensitive ICs can be damaged with no external signs. I once accidentally connected a too high supply voltage to a Freescale iMX6 processor. It mostly worked afterwards, but a few of the interfaces were broken. Link to comment
mansr Posted March 27, 2019 Share Posted March 27, 2019 20 minutes ago, Jud said: Cheap data protection: - You can do RAID, or if you have more than one computer in the home, you can hook up an external drive to each, where they can be both music storage drives *and* backup drives for other valuable data if your computer ever decides to die. Then you just copy over the music from one drive to the other. RAID protects against failure of individual drives. If you connect the wrong power supply (or the right one blows up), it may fry all the drives. 20 minutes ago, Jud said: - Depending on the size of your music collection, SD cards may make a nice, reasonably inexpensive third backup, and are great for taking your music on trips. Do not use SD cards for anything other than temporary storage. They are too unreliable. 20 minutes ago, Jud said: - Backblaze offers unlimited remote backup/storage for around $60/yr. Cloud backup is always a good idea. Link to comment
mansr Posted March 27, 2019 Share Posted March 27, 2019 24 minutes ago, Jud said: The Backblaze pricing is hard to beat for collections of 1TB or more. The disadvantage is that you're restricted to their backup app (which hasn't been problematic in my usage, just works in the background with very low resource use), which is only available for Windows and Mac. For more money, Backblaze, Amazon, and others offer price-per-storage-size options that work with familiar backup tools on a variety of OSs. (I like the ideas behind tarsnap, but it is cost-prohibitive for music storage on the scale most folks here are talking about.) I'm currently using Google Drive with the 2 TB option. I ditched Amazon Drive after they locked out all the clients I was using for accessing it from Linux. Jud 1 Link to comment
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now