|
I just use Illumos with ZFS, like Sun intended us to. Once it's set up and working there is basically nothing to do, and I have a separate box for virtual machines, so all it does is serve me iSCSI LUNs and occasionally get rebooted for kernel patches. Napp-IT makes it super easy to manage the thing, and I never really use it for any of the various UNIX-y things you can do, like BIND or mailserv or whatever. Also, the zfs based boot zone thing is amazing, so many horrible fuckups on my part reversed via mashing the reset button and packing the previous one.
|
#
?
May 29, 2017 17:02
|
|
|
|
| # ? Apr 28, 2024 21:20 |
|
|
So many different operating systems, so many opinions. Can anyone make a recommendation for a good os / setup for 4x4tb hdds serving standard medley of media? I'd something simple, fast and stable, I have never used FreeBSD, and have only used Linux a few years back but I won't say I'm terribly good with a command line. I know windows well enough and all my machines are OSX. The data isn't irreplaceable but isn't something it would be a giant pain in the rear end to replace. I'd like to backup with something like crashplan in addition to external drives kept offsite. I was leaning toward freenas due to zfs and it's simplicity but apparently lots of people online are recommending against it right now. I'm open to going with FreeBSD or some variant of Linux if it's pretty idiot proof / simple enough to fix if / when stuff breaks. Also was wondering if drivepool / unraid were any good, and if I should be honestly concerned with not using zfs / bitrot? I've done tons of reading online but I'm thinking I'm more confused now than when I started. Really, I'm not too concerned about money, I'm not looking to tinker, I just want something simple that just works after I set it up and is reliable / easy to use / has good performance.
|
|
#
?
May 30, 2017 00:54
|
|
|
Bonobos posted:So many different operating systems, so many opinions. Can anyone make a recommendation for a good os / setup for 4x4tb hdds serving standard medley of media? I'd something simple, fast and stable, I have never used FreeBSD, and have only used Linux a few years back but I won't say I'm terribly good with a command line. I know windows well enough and all my machines are OSX. The data isn't irreplaceable but isn't something it would be a giant pain in the rear end to replace. I'd like to backup with something like crashplan in addition to external drives kept offsite. The recommendations against FreeNAS are mostly toward the wholly unliked Corral release (version 10). You can opt to use the current stable, 9.10 or the 11 RC and you'll be fine if you opt to remain with FreeNAS.
|
|
#
?
May 30, 2017 00:57
|
|
|
Bonobos posted:So many different operating systems, so many opinions. Can anyone make a recommendation for a good os / setup for 4x4tb hdds serving standard medley of media? I'd something simple, fast and stable, I have never used FreeBSD, and have only used Linux a few years back but I won't say I'm terribly good with a command line. I know windows well enough and all my machines are OSX. The data isn't irreplaceable but isn't something it would be a giant pain in the rear end to replace. I'd like to backup with something like crashplan in addition to external drives kept offsite. Windows with stablebit drivepool. Might be right for you.
|
|
#
?
May 30, 2017 01:02
|
|
|
I've decided my next file server is going to be a Debian system running an mdadm raid-6 (five 8TB drives to start with, expand from there as needed). Any gotchas I need to look out for when setting this up?
|
|
#
?
May 30, 2017 07:37
|
|
|
ZFS isn't simple at all if you want to expand later on and is total overkill for him Linux ISO storage. My recommendation as always is unRAID if you want to pay. If you want free then Go FreeNAS I guess (not 10.x). E: there's an unRAID demo if you want to try before you buy.
|
|
#
?
May 31, 2017 19:22
|
|
|
I'm not interested in either ZFS or UnRAID. Edit never mind looks like you were responding to Bonobo
|
|
#
?
May 31, 2017 22:35
|
|
|
Farmer Crack-rear end posted:I'm not interested in either ZFS or UnRAID. Yes. Sorry, you slipped in there when I was responding. Congrats on your 8TB drives!
|
|
#
?
Jun 1, 2017 01:29
|
|
|
Well, I just lost my server to that intel avoton defect. Important data is all backed up and the drives (and boot drive) are fine for now. The machine just died as I was using it (BAT sensor on the board reported 0 volts) and then after pulling the plug and reconnecting it it would not boot, give any video output, or anything I can work with. Luckily I have an old desktop with just enough SATA ports to make this work temporarily. It's probably better to just grab new drives at the same time as a new server, these are two and a half years old already. What's the go-to recommendation for a mini-itx board with ECC support? e: Thinking about it the best option is probably to go with the cheapest avoton board, take the opportunity to upgrade to 6 or 8TB disks, and just buy the replacement to the avoton series when it comes out. Desuwa fucked around with this message at 14:38 on Jun 1, 2017 |
|
#
?
Jun 1, 2017 13:03
|
|
|
Avaton's replacement, Denverton, is scheduled to come out during summer - so it may be worth repurposing old hardware until that's an option. For what it's worth it though, even if the CPU died during an instruction, it shouldn't mean that any data has been lost.
|
|
#
?
Jun 1, 2017 14:48
|
|
|
D. Ebdrup posted:Avaton's replacement, Denverton, is scheduled to come out during summer - so it may be worth repurposing old hardware until that's an option. I've got the drives working on my old desktop. They're going through a scrub but yeah, on second thought I don't need to jump the gun just yet and can wait. None of my HDDs are showing any signs of failure, though right now I'm doing them no favours and I need to get some 5.25 -> 3.5 adapters to mount them properly.
|
|
#
?
Jun 1, 2017 15:11
|
|
|
From what I remember, Denverton is looking pretty interesting - Anywhere from 2 through to 16 cores (not threads), clock speeds around 2GHz, RDIMM (meaning ECC and a lot of capacity per DIMM) and one vendor has shown off a configuration with breakout connectors for up to 16 disks, dual 10Gbps SFP+ (probably via Intel ELX5x0), dual 1G (probably via Intel I210), a AST2400 IPMI controller, and block diagrams/documentation have showed that there is support for eMMC, MPX, as well as QuickAssist (which should let enable use of SHA256 checksumming for filesystems that support it).
|
|
#
?
Jun 1, 2017 16:33
|
|
|
Huh, looks like a lot of those new Atom boards come stock with a crapload of SATA ports. Any reason I shouldn't just run one of those for a NAS build?
|
|
#
?
Jun 1, 2017 20:01
|
|
|
AlternateAccount posted:Huh, looks like a lot of those new Atom boards come stock with a crapload of SATA ports. Any reason I shouldn't just run one of those for a NAS build? Not as far as we know. The Avoton boards were similarly perfect for home NAS builds until the C2xxx LPC bug started killing them. e: Actually my Avoton board was over two years old, I was really pushing my luck. Desuwa fucked around with this message at 20:49 on Jun 1, 2017 |
|
#
?
Jun 1, 2017 20:23
|
|
|
So the Avoton boards should be avoided at all costs at this point?
|
|
#
?
Jun 2, 2017 20:45
|
|
|
Unless you can confirm that there's a fix available for them, yeah.
|
|
#
?
Jun 2, 2017 20:55
|
|
|
Thwomp posted:So the Avoton boards should be avoided at all costs at this point? Intel has a new stepping out. Stuff manufactured recently has the fix, and most reputable vendors will RMA an affected board. https://www.servethehome.com/intel-atom-c2000-series-bug-quiet/
|
|
#
?
Jun 2, 2017 21:08
|
|
|
I've got a Synology NAS that I've had for almost exactly three years now, with some WD drives in there. At what point should I consider proactively replacing drives to get ahead of MTBF or should I just wait until they fail and replace one at a time? It's just N+1 so losing two drives'd nuke it.
|
|
#
?
Jun 6, 2017 05:24
|
|
|
There's no hard rule for when a drive is EOL. Keep an eye on the pre-fail S.M.A.R.T attributes of the disks (better yet, chart them with collectd) - if they start changing (raw) values, it's time to replace the disk(s). Anecdotally, I have 4x Samsung HD204UI disks with over 50k power-on hours and they're not showing any signs of problems - generally speaking, once a disk has passed the initial burn-in (consisting of a S.M.A.R.T conveyance and long test, as well as running e2fsprogs' badblocks on it), they tend to last for a very long time. All that being said, remember the most important part about RAID: it's not backup.
|
|
#
?
Jun 6, 2017 13:19
|
|
|
I've noticed when the SMART data for spinup/startup time is increasing, the disk is probably not doing well.
|
|
#
?
Jun 6, 2017 16:59
|
|
|
Question for ya'll, might be better asked in the Linux thread but it is storage related. How do I prevent Linux (ubuntu 16.10 in this case) from auto mounting and partitioning drives? I have some on a system I want to keep as non system partitions for running medusa traffic on but it seems that the OS just decides to mount/partition them. This didn't happen on CentOS 7 FWIW. These are NVMe drives and seems to affect Samsung and Micron ones especially. Disabled auto mount in disks.app but it still does it, there isn't anything specifying these drives in /etc/fstab or anything.
|
|
#
?
Jun 6, 2017 18:15
|
|
|
Backblaze's data shows that SMART has a low false-positive rate (when the numbers change you have a problem) but has a high false-negative rate (lots of drives fail without SMART data changes). Generally speaking I think 4 years is a good time to start thinking about it, and 5 years is time to proactively change over, but there are no hard-and-fast rules, some drives will live to be 10-15 years old if you let them. As always, be sure you have backups of anything irreplaceable. Paul MaudDib fucked around with this message at 21:58 on Jun 6, 2017 |
|
#
?
Jun 6, 2017 18:34
|
|
|
Paul MaudDib posted:Backblaze's data shows that SMART is very accurate (when the numbers change you have a problem) but has a high false-negative rate (lots of drives fail without SMART data changes). Cool, thanks. I know that RAID isn't backup, so I should probably figure out some sort of offsite backup for the important poo poo.
|
|
#
?
Jun 6, 2017 21:18
|
|
|
What utility should I use for SMART monitoring? I tried some WD Data Lifeguard thingy and it gives me this: Now I have seen such dialogs on the internet and everyone seems to talk about it as if it were natural and obvious but those numbers mean absolutely nothing to me. How do I interpret this gibberish? Or rather, what app puts these magic numbers into a human understandable context?
|
|
#
?
Jun 6, 2017 21:57
|
|
|
Falcon2001 posted:Cool, thanks. I know that RAID isn't backup, so I should probably figure out some sort of offsite backup for the important poo poo. RAID is anti-backup. Even with RAID-5 or something, you are still increasing the chances of a failure and/or downtime (and resilvering gets progressively riskier/slower when you increase drive size, modern arrays can be offline for days to do a rebuild). If your data is irreplaceable, you want at least 2 (ideally 3) copies in two separate places (at your parent's house, or a family cabin, etc), with at least one of those being offline. This gives you protection against your house burning down, or Bitlocker screwing your poo poo up. A cakebox of BD-Rs can store a surprising amount of data, and it's convenient to drop off the new discs. You don't need to follow this for your animes or whatever, but if you couldn't replace it and you'd care if it's gone then you need a backup. I went through a lot of DVD-Rs failing due to bit-rot, so if you do this I recommend (1) burning as slow as your disc will let you, and (2) using a rolling parity system. The idea is you use a tool like DVDisaster to make a recovery file of your first disc, then you make a parity file (say 20%) and include that on disc 2. The most recent one always stays on your computer. Then if you ever need to, you can use the parity file from disc N+1 to recover disc N. If N+1 is also corrupt, then you can "roll back" through a damaged section. You can use disc N+2 to recover disc N+1 and then the parity file from disc N+1 to recover disc N, and so on. Here's Backblaze's review of SMART data. Money shot: quote:Operational drives with one or more of our five SMART stats greater than zero – 4.2% So 95% of drives with warning signs ended up failing (low false-positive) and 23% of failed drives showed no warning signs on SMART (high false-negative). Paul MaudDib fucked around with this message at 22:20 on Jun 6, 2017 |
|
#
?
Jun 6, 2017 22:12
|
|
|
EssOEss posted:Now I have seen such dialogs on the internet and everyone seems to talk about it as if it were natural and obvious but those numbers mean absolutely nothing to me. How do I interpret this gibberish? Or rather, what app puts these magic numbers into a human understandable context? Most of those numbers will never really mean anything to humans. But here are some key ones to look at/for: BackBlaze posted:SMART 5 Reallocated Sectors Count One of the popular rules of thumb is if a drive reports any reallocated sectors, consider it in need of immediate replacement. The others, likewise, recommend replacement in the near future. If you catch any of the numbers growing (ie, 5 uncorrectable sectors on Monday, 28 on Tuesday, etc) then the drive is killing itself and is highly likely to fail in short order. Ones that stay stable for a while may or may not die, but if you can warranty them out, do so. Different vendors also implement slightly different sets of SMART reporting (for instance, at least at one point in time WD and Seagate reported current temperatures in wildly different ways), so you'd want to look up what those particular drives report. DrDork fucked around with this message at 22:46 on Jun 6, 2017 |
|
#
?
Jun 6, 2017 22:44
|
|
|
QNAP updated their software rendering Transmission useless. The updated 64bit version has none of the functionality. Any solutions?
|
|
#
?
Jun 7, 2017 03:11
|
|
|
Revert? Don't update? Unsure--QNAP and other appliance-like units don't often have many options.
|
|
#
?
Jun 7, 2017 03:27
|
|
|
EssOEss posted:What utility should I use for SMART monitoring? I tried some WD Data Lifeguard thingy and it gives me this: Download CrystalDisk (the non anime version) and look at the RAW values.
|
|
#
?
Jun 7, 2017 03:55
|
|
|
What's a good way to clone a partition in Windows? I'm retiring a 1TB drive that's pretty old before it shits the bed one day out of the blue. I got a 4TB HGST. I tried dd for windows like 4 times and I can never get it working. The one time I got anything usable out of it, it copied the partition table too. Not only am I not interested in any more MBR drives, I can't use a 4TB drive without GPT even if I wanted to. The last time I got a RAW partition that was useless. So what's a good way to get a copy of that partition over to the other drive while maintaining its GPT status? It's not a bootable drive or anything special like that, but it does absorb the documents folder to keep it off the SSD.
|
|
#
?
Jun 7, 2017 05:45
|
|
|
Clonezilla or macrium reflect do everything I need on non-Samsung disks. Samsung's migration utility is great for their stuff.
|
|
#
?
Jun 7, 2017 06:03
|
|
|
5436 posted:QNAP updated their software rendering Transmission useless. The updated 64bit version has none of the functionality. Any solutions? Grab QBittorrent? Link
|
|
#
?
Jun 7, 2017 15:18
|
|
|
For people using Synology/XPenology, Synology has released their new Virtual Machine Manager as a beta: https://www.synology.com/en-global/beta/2017_VMM You need 2+ LAN ports and at least one BTRFS volume, since the VMs can't be stored on a ext4 volume. If you've got multiple LAN ports teamed, you'll probably gently caress up your network when you configure VMM. I've been running it for a couple days, not bad. It's qemu/libvirt as you'd expect on linux, performance seems fine. 2D over RDP seems like it might be slower than vbox was, maybe because they're using the default cirrus video mode, which I'm guessing is terrible compared to something like the newer qxl? I don't know enough about qemu to say for sure Only problems I've run into are that my Win10 VM isn't too happy being rebooted (though shutdown/power up is fine), and migrating from vbox to VMM was a gigantic pain in the rear end, and a huge waste of time. But it is possible. At least now I won't be months behind on DSM updates, since phpVirtualBox needs Synology kernel sources to compile against, and Synology is so slow at releasing stuff like that.
|
|
#
?
Jun 8, 2017 06:43
|
|
|
Need to disable Fast Startup in the power options. For some reason Windows' hibernation based boot and KVM don't like each other.
|
|
#
?
Jun 8, 2017 07:39
|
|
|
Combat Pretzel posted:Need to disable Fast Startup in the power options. For some reason Windows' hibernation based boot and KVM don't like each other. Yeah, I tried that, but didn't have any luck. Found another mention the same issue with using the VirtIO SCSI driver, which is what Synology is using since they mount the drives as iSCSI LUNs, so looks like there's a bug somewhere that I'm hitting. They seem to update VMM about once a week or so with quite a few updates, so hopefully one of those fixes it.
|
|
#
?
Jun 8, 2017 12:35
|
|
|
Amazon is ending their unlimited cloud storage offer. I'm posting it here because some people may have their backups there. https://www.amazon.com/b?node=16591160011 FAQ posted:On June 8, 2017. Current customers will keep their existing unlimited storage plan through its expiration date. At the end of their existing subscription, customers with auto-renew turned on and 1TB or less of data stored will be renewed into the 1 TB plan for $59.99 per year. Customers with auto-renew off, or who have more than 1 TB of data stored, will need to visit their account's Manage Storage page to opt in to one of the new plans. TLDR: $60 per Terabyte per year up to a maximum of 30 TB. Over-quota data will be deleted within 180 days. The upside is that assholes like this will lose all of their data: 
|
|
#
?
Jun 8, 2017 14:16
|
|
|
See ya amazon unlimited storage. Oh well, it took months to upload my crap.
|
|
#
?
Jun 8, 2017 14:31
|
|
|
eames posted:The upside is that assholes like this will lose all of their data: Is that the guy who had some script pulling webcam vids from a bunch of sites?
|
|
#
?
Jun 8, 2017 15:01
|
|
|
I hope crashplan doesn't get rid of their unlimited plan! Also gently caress that guy with 1 PB uploaded. Ruining it for the rest of us!
|
|
#
?
Jun 8, 2017 19:24
|
|
|
|
| # ? Apr 28, 2024 21:20 |
|
|
phosdex posted:Is that the guy who had some script pulling webcam vids from a bunch of sites? yeah he was up to 1.8 PB according to his last posting. eames fucked around with this message at 19:41 on Jun 8, 2017 |
|
#
?
Jun 8, 2017 19:37
|
|