|
Thanks Ants posted:If you're buying an enclosure anyway then can't you hook something up to the eSATA port that's already in the Microserver? Apparently the eSATA ports on them only run at half speed unless you do some hacked BIOS flashing and .. not really interested in doing that. Anyways that enclosure comes with eSATA as well so should be no problems there, and I could use a USB3 HBA anyways. Sheep fucked around with this message at 20:54 on Dec 13, 2017 |
![]() |
|
I'm not near a PC to link it but there's an adapter that lets you fit 2 3.5 drives in the 5.25 Bay. I currently have 6 3.5 drives and 1 2.5 ssd stuffed into my n40l which is the same size as an n54l. The power supply handles it fine, if you only needed 1-2 more drives.
|
![]() |
|
Minty Swagger posted:I'm not near a PC to link it but there's an adapter that lets you fit 2 3.5 drives in the 5.25 Bay. I currently have 6 3.5 drives and 1 2.5 ssd stuffed into my n40l which is the same size as an n54l. The power supply handles it fine, if you only needed 1-2 more drives. I remember seeing these used in the past. No idea where to find them now. https://www.amazon.com/gp/aw/d/B000S8B8J6/
|
![]() |
|
So I finally got my FreeNAS install up and running (i7 3770, 24GB, 4x 3tb reds) and I'm wondering what to do for remote access from work, so I can tinker etc. Traditionally I've used RDP with 2FA to my desktop, but now I've got an i7 running 24/7, I don't really want my OC'd gaming machine running too. I was thinking of setting up a VM to use as a terminal server, problem is I don't really have room in the budget for a win 10 Pro licence, and I don't really want to run a full Win 10 instance in a bhyve anyway. I have a lubuntu VM running, but RDP/terminal server options for linux all seem to be dogshit. I can L2TP VPN to my home network via my router (USG), but then I have to pass all traffic through my home connection for it to work. OpenVPN and Softether clients seem to trigger Sophos AV on our work machines, causing it to lock out traffic on Port 80 for some reason so that's out (which is a shame because I had that working real well). I mostly want to be able to access web apps running in Jails and the odd terminal on the NAS itself, and I'm looking for a nice light solution. SSH tunneling might be an option, but I have 0 experience and no idea to even start looking. Or if there's a better way to set up L2TP or something. What are people here using?
|
![]() |
|
Depending on what apps you're using, it might not even be necessary unless you're super security-conscious. Things like Plex, Transmission, etc., all have password-enabled web front-ends that you can expose and access without needing a VPN. Doesn't help you on the console front, admittedly, but it's a lot easier to get working.
|
![]() |
|
It is possible to set up a VPN that only adds a route to your local network, instead of rerouting the gateway as well.
|
![]() |
|
IOwnCalculus posted:It is possible to set up a VPN that only adds a route to your local network, instead of rerouting the gateway as well. Unfortunately the version of the Unifi Controller that's available for FreeNAS doesn't seem to be able to do that, so I'd have to set up an LT2P client in a jail and expose that, which I'm not opposed to doing I guess.
|
![]() |
|
I know in openvpn it's something you can set up client side, but I can't recall what. Which is annoying because I need to do that too ![]()
|
![]() |
|
I had it working in OpenVPN via the OVPN client, but I had to uninstall that, so it's Win 10 VPN client or fuckoff apparently.
|
![]() |
|
Minty Swagger posted:I'm not near a PC to link it but there's an adapter that lets you fit 2 3.5 drives in the 5.25 Bay. I currently have 6 3.5 drives and 1 2.5 ssd stuffed into my n40l which is the same size as an n54l. The power supply handles it fine, if you only needed 1-2 more drives. Unfortunately it's a 4-drive RAID 6 array I'm pulling out to the external enclosure so no dice there. Would be nice to grab a dual-3.5" drive bracket for the upper bay of the N54L but I can't for the life of me figure out what to plug into Amazon to find them. Edit: found one at twice the price it should be. IOwnCalculus posted:I know in openvpn it's something you can set up client side, but I can't recall what. The fundamental concept is called split tunneling, no idea what any specific implementation calls it beyond that but at least now you know what to CTRL+F in the documentation! Sheep fucked around with this message at 03:44 on Dec 14, 2017 |
![]() |
|
Sheep posted:Apparently the eSATA ports on them only run at half speed unless you do some hacked BIOS flashing and .. not really interested in doing that. Use the other bios, it also offers some other things that are well worth it. Having all you sata ports running on 6gbps for example. Had mine running the hacked bios for over two years. No problem whatsoever.
|
![]() |
|
Don Dongington posted:So I finally got my FreeNAS install up and running (i7 3770, 24GB, 4x 3tb reds) and I'm wondering what to do for remote access from work, so I can tinker etc. I'm familiar with using a SSH tunnel from remote machines into my home network. I'm assuming your work machine is Windows and I've used Firefox In the past for tunneled web traffic b/c it had independent network proxy settings from the OS.
If you can SSH into your lubuntu instance from your work machine you should be able to then SSH into your FreeNAS box. If you end up putting the plink SSH tunnel command(s) into a script/batch file, I recommend setting up public key authentication so you can login to your lubuntu instance from your work machine w/o having to type in or embed your password in the script. ---- 1 I believe DNS lookups will still happen on the your local (work) network unless you configure things differently. E.g., if your company is preventing DNS lookups to https://www.twitter.com you won't be able to access it through the SSH tunnel unless you alter your DNS settings too. Perhaps there's a Firefox add-on for that as well dbcooper fucked around with this message at 10:24 on Dec 14, 2017 |
![]() |
|
Hello people, Im an amateur, but capable FreeNAS user. Im on 11.0u4 right now, and I want to upgrade to 11.1. I virtualize FreeNAS, so Im thinking of doing the following: 1. Save FreeNAS configuration. 2. Power down FreeNAS 11.0u4 VM. 3. Configure a new VM with identical specs to my u4 VM. 4. Boot new VM with the 11.1 installer and install it; then load my configuration backup Im pretty sure this will work just fine to setup 11.1, but heres my burning question: if for one reason or another 11.1 sucks, will I be able to roll back to 11.0u4 ok? By that I mean, will my ZFS pool have any problems going from being used with 11.1 to being used with 11.0u4?
|
![]() |
|
Is there a reason you don't want to use FreeNAS's built-in upgrade system? Even as a VM it should be able to take care of itself without issue, though saving the configuration before an upgrade is never a bad idea. As for rolling back, yeah, it's easy: through the GUI, go to System -> Boot and simply select an older setup and click Activate. If you can't get into the GUI for some reason, when FreeNAS is booting you get a little boot menu for a few seconds where you can do the same thing. Your pools will be fine.
|
![]() |
|
DrDork posted:Is there a reason you don't want to use FreeNAS's built-in upgrade system? Even as a VM it should be able to take care of itself without issue, though saving the configuration before an upgrade is never a bad idea. I could do that, but I figured keeping them separated would be easier. To be frank, Im less concerned about the pool and more concerned about the jails I have. They exist on the pool, and therefore should be fine, right?
|
![]() |
|
If you use FreeNAS's built-in upgrade system, yes, they will be just fine. Unless you're trying to do something very specific/fancy, there's not a whole lot of reason to do completely separate installs for a dot release. I could see if you were trying to check out the differences between 9/11 and Coral or something, but this is nothing near as substantial. Rolling back through the GUI/Boot loader is about a 5 minute process total. I have never had much luck getting jails/VMs to gracefully move around from one instal to another, though. The jails do live on the pool, so you shouldn't lose data, but having to re-spin them up is a hassle.
|
![]() |
|
DrDork posted:If you use FreeNAS's built-in upgrade system, yes, they will be just fine. Thanks for the feedback. I'll take the plunge sometime this week! Maybe I'll create a snapshot in ESXi just to be sure.
|
![]() |
|
So I was thinking of building a combo HTPC and NAS. But the HTPC thread recommended just getting a Shield TV and a separate NAS. I already have a good 5TB worth of data I need to store, with plans to expand that in the future obviously. Looking for some redundancy too. Should I look at pre-built solutions? It seems like rolling my own would be cheaper. It's been almost a decade since I used linux but I remember it being pretty easy to install and get running. What hardware do you guys recommend if I build my own?
|
![]() |
|
MonkeyFit posted:So I was thinking of building a combo HTPC and NAS. But the HTPC thread recommended just getting a Shield TV and a separate NAS. I already have a good 5TB worth of data I need to store, with plans to expand that in the future obviously. Looking for some redundancy too. Should I look at pre-built solutions? It seems like rolling my own would be cheaper. It's been almost a decade since I used linux but I remember it being pretty easy to install and get running. What hardware do you guys recommend if I build my own? Linux is easier than ever, since there is now a lot more how-to documentation out there. Linux now has good ZFS support/etc. 5 TB is within the range of a single modern high-capacity disk. BestBuy has been running EasyStore 8 TB disks for $140-160 for a while now, and they have a WD Red or a white-label Red inside. Depending on what you want, you could set yourself up with a 2-bay or 4-bay NAS and leave yourself future expansion room. For rolling your own, I think the U-NAS cases are pretty cool. They make 2/4/8-bay enclosures that are about as large as a Synology, but for the same price you can give yourself a lot more power and expansion room, plus ECC. I am going 8-bay with a dedicated NAS, I went with a SuperMicro mATX LGA1151 board with 8x SATA ports and quad NICs, with a 7100 and ECC, but if you have heavy transcoding needs or larger RAM requirements you could think about a quad-core Xeon or a Denverton board. If you want HTPC in the same box figure on buying a model with an iGPU (any consumer processor, although you'll give up ECC, and I think the Xeons with iGPUs usually end in 5) or using a low-end discrete GPU for handling outputs (eg GT1030). And look up your decoder requirements before buying. Paul MaudDib fucked around with this message at 12:06 on Dec 16, 2017 |
![]() |
|
Speaking of ECC for NASes, hasn't anyone figured out to build a cheap Ryzen ECC build yet? How are the low end motherboards holding up with ECC? Low cost low power ECC Ryzen APU build could trigger my itch for a new build.
|
![]() |
|
yomisei posted:Speaking of ECC for NASes, hasn't anyone figured out to build a cheap Ryzen ECC build yet? How are the low end motherboards holding up with ECC? Low cost low power ECC Ryzen APU build could trigger my itch for a new build. If you care enough for ECC why are you settling for an unvalidated platform let alone one with a known problem when you put it under load?
|
![]() |
|
You don't need ecc *runs away*
|
![]() |
|
I ran ECC RAM on a consumer AM3 board a few home server iterations ago. I probably didn't need it, but it certainly didn't hurt, and it was significantly less expensive to pay the small premium on the DIMMs without the Xeon motherboard and CPU tax. Then I ended up reusing the ECC DIMMs when I got a good deal on a used Xeon motherboard and CPU ![]() Ryzen memory controllers have ECC functionality, so it should work as long as the traces for ECC-related pins exist on the motherboard, and it's probably a step up from non-ECC memory even though the platform as a whole is not validated for ECC. The only AM4 motherboards I've seen that definitively have ECC DIMM support are on the higher end with an X370 chipset. Googling for specs and trip reports is probably the best you can do to determine if a particular motherboard will do it pre-purchase. I guess what I'm saying is get it if you want it, or don't if you think it's unnecessary or need a validated platform. ![]() P.S. If your Ryzen CPU exhibits problems (i.e. segfaults) under load, then it is defective and AMD will replace it under warranty with few questions asked. That the whole product line/architecture is bad and should be avoided is FUD. SamDabbers fucked around with this message at 21:25 on Dec 16, 2017 |
![]() |
|
The real use case for AMD ECC is lower / mid range. Intel forces you up to high end if you want ECC and full validation.
|
![]() |
|
Looking for a backup solution for most of my music/movies/pictures/config files. Not looking to stream anything but music. Right now I'm considering a DS216se. For my uses would there be any downside to buying something so weak?
|
![]() |
|
IOwnCalculus posted:The real use case for AMD ECC is lower / mid range. Intel forces you up to high end if you want ECC and full validation. ECC is officially supported on Pentiums and i3s, I don't know if you consider that "full validation" but they are the exact same chips coming off the same lines as Xeons, there is no realistic case where one would work and the other wouldn't. In contrast, literally nobody even tests ECC on AM4 platforms at all, and the handful of boards that purport to support it have had it break in various BIOS versions, because it's not an official feature. Also, again, until AMD fixes their memory controller ECC support is pretty meaningless. Segfaults are just the tip of the iceberg, it also appears to have some more subtle memory corruption going on, judging by the way that ASLR causes some crashes. The IMC is just flat-out glitchy. If your processor is stable for a 24h+ burn test of kill-ryzen then it's reasonably safe for consumer use, but it doesn't prove the fault is gone entirely, it's just an asymptotic reduction in the fault rate. You simply can't trust first-gen Ryzen to be fully stable under load, regardless of the quality of an individual sample. There is a reason that the enterprise chips (Epyc) are being released on a new stepping. The platform just is not trustworthy until they can resolve that bug (a new stepping is coming to the consumer platform in a couple months). It certainly doesn't make sense to put ECC on it given the known issues - ECC doesn't fix the IMC telling it to do dumb shit. Paul MaudDib fucked around with this message at 04:37 on Dec 17, 2017 |
![]() |
|
Paul MaudDib posted:ECC is officially supported on Pentiums and i3s, I don't know if you consider that "full validation" but they are the exact same chips coming off the same lines as Xeons, there is no realistic case where one would work and the other wouldn't. Ryzen dies are identical to those in Threadripper and EPYC. Should those platfroms not be trusted either since they have the exact same IMC?
|
![]() |
|
SamDabbers posted:Ryzen dies are identical to those in Threadripper and EPYC. Should those platfroms not be trusted either since they have the exact same IMC? No, Epyc is actually on a newer stepping for this precise reason (B2 vs B1). The enterprise community isn't going to put up with that shit. For some reason they never pushed this stepping down to the consumer chips. Threadripper is on B1 so it's theoretically susceptible to it, but it's cherrypicked silicon, the segfault bug (really, just a glitchy IMC) is inherently a manufacturing problem and the highest-quality silicon is relatively unlikely to manifest the issue. It's also possible that whatever NUMA they've layered on top avoids the trigger condition for the bug. A whole lot of the Ryzen silicon is bad. The vast majority (>95%) of all early chips have it, and a significant fraction of the current-production chips still have it. You really should not use first-gen Ryzen for anything production critical, just wait another couple months for the Zen+ stepping. If you're OK with ZFS segfaulting under load then go hog wild, but unclean termination is not good for your filesystem, and it definitely doesn't make any sense to put ECC on a system like that. The fact that ECC is not even a supported feature on consumer chips is just the cherry on top. The board partners are not testing it and you should absolutely do your own testing each time you update BIOS/microcode to be sure it's actually working. Threadripper is fine in my book though - ECC is an official feature there and it is getting tested. In practice it doesn't seem to have the segfault bug either, or at least you are buying your way to the silicon that is the least susceptible to the issue. If you want to be maximally paranoid though, TR is a niche platform and really hasn't seen the amount of battle-testing that LGA115x chips get, and the absolute safest move is to give first-gen Ryzen a pass entirely until they get a new stepping for the consumer chips. Paul MaudDib fucked around with this message at 05:16 on Dec 17, 2017 |
![]() |
|
The other thing to bear in mind with Epyc is that there are only a handful of systems in the wild so it's possible that B2 also has a faulty IMC. There really hasn't been extensive testing on volume-produced samples since it hasn't ramped yet. The Epyc paper launch happened 6 months ago, which was the same time as Michael Larabel found the segfault bug, and even today there are still no systems available, basically just engineering-sample quantities. Nobody but Tyan and Supermicro has any systems available period, and Supermicro is rumored to have shipped only a few hundred total, nobody could get chips from AMD until this month and nobody knows why. It's entirely possible AMD saw this bug and decided to push back volume production until they could get a new stepping out. Because again, enterprise customers won't put up with this shit.
Paul MaudDib fucked around with this message at 05:18 on Dec 17, 2017 |
![]() |
|
Also enterprise validation is a thing, and getting Tier 1 server providers a fully validated system takes a crapload longer than telling board partners 'here's the spec, go nuts'. HP just recently announced the availability of servers that use Epyc. And yeah, chances are the B2 stepping fixed a lot of errata they are declining to state at this time, specifically because all the first gen Ryzen chips are affected. I have a 1950X in my garage with 64GB of ECC, and it's working well. I'll have to run the kill-ryzen script on it at some point to see if I can get it to fault out or throw ECC errors.
|
![]() |
|
I'm becoming increasingly interested in running a machine with NIC removed as a secure store of data. It doesn't have to be huge capacity, but I'd like desktop Linux on it. Preferably Fedora/CentOS but Ubuntu would do. Is it feasible to run something like an Intel NUC with no wireless card in it and the ethernet disconnected, running a small ZFS mirror and not have it turned on 24/7? Would this negate the effect of ZFS? I'd like an offline PC running Linux that's guaranteed to retain data integrity down to checksum level, basically. And if a file checksum fails then I'd like it to tell me. Would it be better to leave it permanently powered on if I'm using ZFS? EDIT: Is it feasible to run a NUC with two identical USB sticks plugged in as a ZFS mirror? Then if one stick fails a scrub it gets replaced? Speed of data retrieval isn't majorly important in my use-case. apropos man fucked around with this message at 07:02 on Dec 17, 2017 |
![]() |
|
Methylethylaldehyde posted:Also enterprise validation is a thing, and getting Tier 1 server providers a fully validated system takes a crapload longer than telling board partners 'here's the spec, go nuts'. HP just recently announced the availability of servers that use Epyc. Yeah, they announced availability, but the server lines themselves have been announced since June. There was no availability because there were no processors to put on them, the boards themselves have been ready to go for a while. Partners have been trying to get chips and couldn't get them. Methylethylaldehyde posted:And yeah, chances are the B2 stepping fixed a lot of errata they are declining to state at this time, specifically because all the first gen Ryzen chips are affected. The real question for me is when AMD knew they had a problem. They may have known in a generic "our IMC is running really close to tolerances and is not particularly stable" sense earlier, but did they know that some of the chips were outright failing and ship anyway, or was that a surprise to them in June? B2 has been around a while and it might have the issue too. I'll be real curious whether production Epyc ships on a newer C1/B3 stepping, and/or if it has the segfault issue. If they're shipping on B2 and it's stable there may be some heavy binning involved - the alternative being that AMD knew earlier and B2 has a fix (a tuned up memory controller). There is some thought that B2 might actually be pretty similar to B1, just basically the designation for the mirror-image die that they're using in 2 of the 4 positions. We don't really have a list of errata yet. I think they would want to push at least some errata fixes for the enterprise crowd, but the question is why they wouldn't push those back to the consumer chips ASAP. They have continued manufacturing Ryzen on B1, which is circumstantial evidence for the "mirror image" theory. Methylethylaldehyde posted:I have a 1950X in my garage with 64GB of ECC, and it's working well. I'll have to run the kill-ryzen script on it at some point to see if I can get it to fault out or throw ECC errors. Like I said, Threadripper is unlikely to have the issue (AMD claims it's the best 5% of Zeppelin dies) but I don't think it's categorically impossible that some samples may have it. Another task that I've heard semi-reliably reproduces the issue is x265 encoding. Paul MaudDib fucked around with this message at 08:04 on Dec 17, 2017 |
![]() |
|
Xeom posted:Looking for a backup solution for most of my music/movies/pictures/config files. Not looking to stream anything but music. Right now I'm considering a DS216se. I haven't "streamed" music as much as I have played the files back from the network location, in a music player with a local database, which works fine even on a DS112j. I guess if you're serving to more than one or two users simultaneously, even file transfers are going to suffer, probably. But it's not inherently unusable. The step up to a DS216j is relatively inexpensive and significant though.
|
![]() |
|
apropos man posted:I'm becoming increasingly interested in running a machine with NIC removed as a secure store of data. It doesn't have to be huge capacity, but I'd like desktop Linux on it. Preferably Fedora/CentOS but Ubuntu would do. bit of a waste of a computer that. could just run a proper file server with freenas or something with all the zfs checksumming niceties which you periodically back up to some external drives or something
|
![]() |
|
Just buy a cheap laptop to store your buttcoin
|
![]() |
|
Has the ZFS/ECC RAM thing been debunked now? I can't remember what the general consensus was the last time I read up on it. Sorry if I'm opening up that can of worms again ![]()
|
![]() |
|
apropos man posted:Has the ZFS/ECC RAM thing been debunked now? If you're not doing white box VM stuff, it's perfectly fine to ignore ECC. It's a nice to have feature but definitely not needed for Linux ISO's. You should be backing important shit off your NAS anyway.
|
![]() |
|
Matt Zerella posted:If you're not doing white box VM stuff, it's perfectly fine to ignore ECC. It's a nice to have feature but definitely not needed for Linux ISO's. You should be backing important shit off your NAS anyway. Yeah. I thought the argument for ECC had been put down to pure zealotry. I think I'm gonna automate some snapshots this time round. I run ZFS on another server that's constantly accessible by VPN but I haven't bothered with ZFS snapshots on that one. I think I'll use a combo of snapshots and pulling a borg repo off of it and pushing that up to S3 storage.
|
![]() |
|
Backups > ECC Use ECC for data you care about, but only after you have good automated backups in place.
|
![]() |
|
When I saw the WD 8TB Easystore deal at Best Buy last week, I bought 5. I don't need that much storage but figured it was better than spending the same amount of money for half the storage (4TB drives). This is my first personal NAS but I'm comfortable with Linux, Linux servers, PC and server hardware, mdam, etc. I have my [old] PC from 2010 (Intel i5-750, Gigabyte GA-P55A-UD4P and Antec P183 case) than I'm not using. I was planning on setting up the drives in software RAID as the motherboard has enough SATA plugs onboard to support them. I've never used ZFS but it sounds interesting and I'm thinking ZFS-2 instead of RAID 6. My use cases are:
Oh, I'd also like to consolidate all my old documents and pictures once they're all stored on a reliable single system. Ideally this would be a one time thing. Any OS suggestions? FreeNAS? Raw centos or Debian? Some paid product? So I don't feel I need high throughout or the ability to run VMs on it. If I can get away with it, maybe USB sticks for the OS? Once I offload some data I should have two or more (old) 1TB HDDs I could do something with in the NAS. Edit: Added old PC parts/links dbcooper fucked around with this message at 22:37 on Dec 18, 2017 |
![]() |