Update on Wendell's Server Build for GN: Projects & Capacity

[@addisonmartin730]

I really enjoy this kind of server content

[@BryceAC]

Nice teleporter

[@AdmireFiestaGuild]

I like how the server is professionally mounted on some IKEA stuff in a 19" rack :D

[@gnsteve8846]

It's a printer table my dad built years ago. Might as well keep it in service, it just needs to hold a box.

[@jonesconrad1]

well there's not much point mounting it in that rack. it's probably safer on the table.

[@killtyrant]

@@gnsteve8846 Here here for God building a table for jesus

[@aemonblackfyre4159]

KillTyrant wasnt josef a carpenter anyways? :D

[@OFDM-network]

Its called price efficiency :)

[@positivemelon7578]

Keep in mind that even after this upgrade ZFS will probably still use all that memory. That's good, otherwise the memory is wasted. Memory may appear as almost full but when you start using more memory, for example by running VMs, ZFS will start using less memory.

[@steven44799]

the way ZFS works, it will use all the memory it can get as a cache. i have seen ZFS using 300+gb of ram for cache.

[@franzpleurmann2585]

You can set a limit with zfs.conf and zfs_arc_max.

[@Multimeter1]

This is what he’s probably experiencing.

[@Masterrunescapeer]

@GNSteve don't use a VM, rather run a docker image like adriel/h265ize from docker hub. Just keep it running there, it will also move your files to a different folder if required. Because it's a docker image, you don't need to dedicate memory per container, you can also spin up multiple docker images for different folders if they're not all in the same root folder (image I recommended preserves file structure). I suggest using kitematic if you need an easier GUI.

[@nerdedkyle3515]

Really liking these side channel vids

[@simoSLJ89]

Great Steve. Happy to see this second channel grow with bike stuff, behind the scenes stuff..and maybe cats! :D

[@kakarroto007]

I like that you even thought of having a constant "local" backup of the server.

[@Biggarou]

Great to see this other channel up and running. But please steve don't burn yourself out in work.

[@faffabout9412]

Hey Steve, Network Engineer here, As long as you have a good retention plan, you can recover from almost anything. Biggest things to consider 1. The main goal of RAID is redundancy, its not so much to protect data, rather its to protect the server and prevent a catastrophic failure that cost you a ton of work to repair, for a company like yours I usually use Raid 5, 6, or 10. Raid 6 if availability is the main concern, 10 is like 6 but has some data savings, usually use that in a SAN environment 2. Your backup solution should avoid using UNC file paths, at least one of your backups should use an agent or tunneling system to copy files, and should come with an aggressive retention policy this is mostly for security, as many crypto attacks can write to UNC file path backups, have seen too many companies that thought their old Symantec was fine, then it wasn't and I had to tell them their only option was to pay the ransom or pay to have a third party attempt decryption. Another key reason for this, is that often crypto payloads are on delay, or a hacker who already has compromised your network. Will figure out what type of backup you are using and attempt to circumvent it before they release the payload and encrypt your data. That's where your retention policy is important, if you can narrow down when they got access, you can restore back to the earliest point before they gained entry, provided you have a restore that is close to that timeframe. Thanks again for the content, always appreciate your vids

[@JmbFountain]

At work, we have monthly backups on tape going back over a decade. This is likely Overkill, but storing backups on an offline medium is the best protection against ransomware.

[@raulsaavedra709]

Raid 5 has been NOT recommended for some time now on physical disks. The issue is that whenever one disk fails, recovering requires a lot of activity from the other disks in a RAID5 array, and likelyhood of another failure goes up precisely because of that, right after you had already one disk failing. So RAID5 in a way defeats the design purpose of RAID 5 in the first place, can render the entire array unrecoverable. Cost of full redundancy is a lot lower nowadays, so RAID 10 is a much safer and recommended choice over RAID 5.

[@PainterVierax]

@@raulsaavedra709 not nice to copy-paste your comment.

[@raulsaavedra709]

​@@PainterVierax I think replies to comments disappear in YT if the original comment gets removed. Wrote the original comment directly on the video for GN's sake. My similar comment to your comment here was just for you to check out

[@PainterVierax]

@@raulsaavedra709 "for GN's sake." yeah, I'm sure he's already aware of that.

[@LindsayDunia]

3-2-1 is a great backup model. You can google 3 2 1 backup model.

[@_BRawlR_]

I've always done that for clients (came up with it on my own) but didn't know it was a "thing"

[@nojoojuu]

Saving all the videos must take a huge amount of space... ? One backup or 2? Loving these vids!

[@steven44799]

your backup strategy/plan is fairly solid. i do local, remote (to another property we own/lease via VPN) and remote (cloud with infinite version history). the intention is that unless everything has proverbially hit the fan or you need something that was versions earlier, the cloud backup is your last resort as you have the least control over it and the most restricted access. if you have to you can pickup your remote solution and move it to the office to get up and running while you restore over 10+gbit. being linux/unix to linux, a cron job to rsync from the office to home is likely the simplest solution, if you dont want to setup a site to site vpn between them you can rsync over SSH (though im not sure if synology supports that), but you need to open your remote end SSH port (use a non standard port number (nat port forwarding from 61834 to 22 for example) and restrict access by IP if you can (either at the port forwarding or OS level)).

[@Ghyus01]

You guys inspired me to do my own unraid setup and I love it! Not sure if it applies but you can get a docker plugin that has a watch folder where it'll automatically transcode any files dropped in it. Might be of use!

[@kennyj4366]

Thank you for the video content Steve. After you install the Ram and run your VM I hope that means another video. You probably already know this but ZFS uses what ever memory is available..

[@blackmennewstyle]

Pretty sure Linode, your favorite server hosting is offering remote backup service :) Nice setup guys, i'm actually impressed on how quiet everything is :O Have a great weekend and keep it up the great job!

[@redsquirrelftw]

Online disk space is way more expensive than in house. In house you pay for it once, online you pay per month. For example adding 1TB of disk space to an online server will bump you up like 20-40/mo extra.

[@cjsfriend2]

Great video, I enjoy all the server content!

[@fr05ty]

@GNSteve I'm not the smartest at this but you could set up a handbrake docker container then setup up a custom preset in handbrake then you ad that to the docker containers parameters `Automatic Video Converter Preset` and set the `Watch Directory` and `Output Directory` for the the container, to use for auto converting video. Then if someone else knows more about scripting on unraid you might be able to move files into the auto convert watch folder to have them proccesed, and just so you know if you go to unraids web gui > setting > CPU Pinning, you set handbrake to only have acces to xx amount of cores (eg. core 3,4,5,6 of 8 cores) so it doesn't hog all your cpu power. Core 1 and its hyperthread or smt are what unraid uses so its usually best not to assign to much to that. if you didn't do any scripting you could use krusader (another docker container, just be careful with this its direct file access on the server but you can set a VNC_PASSWORD in the container) to move the files to the watch directory manually it will be a lot quicker to move files between directory's this way that over the network

[@MrInsanityflea]

For a small office, that setup is great! You don't need any more redundancy.

[@redsquirrelftw]

I would look at getting a 4 post rack as it will give more room to expand. Server racks are kinda like a drug though, once you get one... you get two.... and you start to fill them. I got a video on my setup. All my stuff is old by today's standards though but it does what I need.

[@FreshEggs]

brilliant video, lets see more such as installing a new 42u rack and move everything over etc... :)

[@ruiamaral4669]

The memory will always run low because that's the way zfs works and that's ok. It will always use as much memory as you can throw at it. Having said that. More is always better.

[@davkas19]

I don't know if you said what OS the black case was running, but cloudberry backup client backing up to backblaze is a good option. You could also setup a VPN between the office and your house, and backup to the Synology at the same time. Cloudberry calls it a hybrid backup. (local and cloud at the same time)

[@amp888]

Your disk shelf sounds quieter than mine does while running (I've got the same NetApp DS4246, plus a DS2246). You didn't do any fan mod to it, did you (I didn't see any in the build video, I don't think)? I've seen a couple of videos of people replacing the standard fans, but it's a destructive process AFAIK (requires cutting and splicing/resoldering wires).

[@gnsteve8846]

It's possible that Wendell did something to it (or its previous owner) before we got it, but as far as I know, there are no modifications on it. It's actually not too loud and mostly blends in with the other white noise. Our disks aren't running against the limits or anything, so maybe that contributes to it. I'm not sure how those fans work, but if they spin-up based on internal case temperature, maybe ours is just running cooler since it's less populated? Not sure!

[@GeoffSeeley]

Just a thought Steve. Maybe run a server class machine in a server class chassis? Yes you'll still run into fitment issues but they'll be a LOT less than trying to cram enough hardware for a server into a SFF case.

[@Lead_Foot]

With ZFS your ram is l1arc(basically read cache). It is possible to limit how much memory it can use if you want to save memory for vms. For a cpu cooler I'd recommend an ID cooling is-60 with an nf-a12x15 if you have low profile memory. It's 55mm tall and that case takes 57mm.

[@ewenchan1239]

I know that the initial cost is quite high, but get yourself at least a LTO-7 SAS (or Thunderbolt if you want to use YoYotta and a Mac) to back up your stuff. For the amount of data that you'll be backing up, it's actually not going to be too bad for you. I have LTO-8 at home now and ten 12 TB tapes (so I can back up to 120 TB of data, which I will pre-compress using 7-zip) and it's already been immensely helpful to me. If you want to make it as simple for yourself as possible, get yourself a LTO-8 Thunderbolt tape backup drive and you can just write two copies of each tape that you backup -- where you can keep one set of tapes at your house, so that like you said, if something happens or fails, you can just run home and grab the tape. I STRONGLY recommend that.

[@JmbFountain]

You pre-compress using 7zip? For compressed writing to tape I would always recommend a .tar.gz

[@ewenchan1239]

Johannes "Fountain" Micha Yup. 7-zip, using the LZMA2 compression algorithm, with the maximum compression level (cf. sevenzip.osdn.jp/chm/cmdline/switches/method.htm), produces the smallest archive file, and the fact that it supports upto 32 threads by default, whereas, I don’t think tar nor gzip supports that. The largest, single 7-zip file that I have written to my tape is 4.4 TB, compressed using those settings. You can test it yourself in terms of time and resulting size (e.g. compression ratio). (And of course, compression ratio/size, and speed depends significantly on the type of files you’re compressing.) I have found that rather than relying on the tape drive to compress the files for me, it is better and faster overall for me to pre-compress my files using 7-zip prior to writing to tape. It’s unfortunate that only Bzip2 supports multithreaded compression and decompression, but Bzip2 also has the limitation that it creates a bz2 file for every file you’re compressing, so that’s not very useful/practical, generally speaking.

[@DigitalMoonlight]

@@ewenchan1239 You can group the files by taring them and compressing with bz2 using tar. Not really a practical solution except in extremely memory starved systems where you need more compression than gzip but don't have the memory for lzma based solutions like 7zip and xz. lzma decompresses much faster than bz2 and I think that greatly outweighs compression time since I'd rather a smaller backup that compresses slower but doesn't sacrifice decompression time too much. If you are looking for a tar based format as a replacement for 7zip files (not really sure if there's a practical difference when using tape) you can always try using the xz or lzma compression flags with tar. I hope you and any other readers find this info mildly interesting or informative.

[@ewenchan1239]

@@DigitalMoonlight So my thought process when selecting the archive format, and the compression algorithm was I wanted something that will compress quickly, and result in the smallest file size since I was writing to tape. The idea behind the small file size/high compression ratio is so that when the archive is written to tape, it won't take a long, nor take up so much space on the tape, per tape, since I am also writing two copies each. As far as memory is concerned, I wasn't really worried because my compute nodes each has 128 GB of RAM and my headnode has 64 GB of RAM, and I think that the node that's writing the data to tape has at least 32 GB of RAM. (My four compute nodes has a total of 512 GB of RAM between them, and as such, I wasn't really worried about that.) The downside with straight 7-zip is that it doesn't preserve the metadata properties about the file (ownership and date/timestamps, but they tell you how to do that on the 7-zip command line website -- the command that you will need to use if that's what you want to do (e.g. backup a directory). It is somewhat unfortunate that 7-zip decompression doesn't appear to be multithreaded like the compression is, but oh well. And given that LZMA decompression is faster than bz2, and in my experience, most of the time, I can only decompress at around 30 MB/s or so, so...it's not terribly fast for that. But then again, if I am pulling a 4.4 TB 7-zip file from tape, that takes a while, and then having to unpack it also takes a while, so my "normal" turnaround time for a data retrieval request is 2 days and that's just limited by transfers speeds (since even though LTO-8, on paper, is spec'd to a maximum of 300 MB/s, even when writing to a NFSoRDMA RAID0 array, in practice, it's closer to 200-270-ish MB/s). I'm open to ideas on how to achieve BOTH objectives: speed AND size. But if I had to sacrifice something, I would much rather sacrifice speed rather than size.

[@DigitalMoonlight]

@@ewenchan1239 I did a little bit of quick research and it looks like using tar with pixz might fit the bill. Pixz does parallel compression and decompression and the compression ratio should be comparable to 7zip since xz is based on lzma. Might be worth doing some bench marks to see if you get appreciable improvement. Of course this assumes you're running *nix, though I don't see why Cygwin or WSL wouldn't work as well for Windows. I also couldn't find up to date info if they added parallel decompression to the official branch of xzutils but I did see some posts mentioning if the file was compressed with multithread it will decompress with multithread. Link: github.com/vasi/pixz

[@bobxdark370]

Hey Steve, it maybe a good idea to put that server on it's side, so that the hard drives are vertical. In the last 30 years I have managed over 1000 servers for a myriad of companies, and I can guarantee you that when a drive starts failing, it will at least warn you nicely if it's mounted vertically, but if it's horizontal, you will experience a different type of heart attack... more pronounced. The main reason is the wobbling the platters start doing when the bearings start wearing, and specially in a box filled with drives, that causes periodic vibration resonance that can kill several drives at once.

[@lucappelman8558]

You should take a look at the tdarr docker image. It can automatically transcode video files with ffmpeg or handbrake. It's really easy to write your own plugins for it too!

[@sparkyenergia]

ZFS Mirror and ZFS Mirror Split. You create a three drive mirror ( 3 * 1tb HDD = 1tb capacity). Each drive is a perfect copy of the others. Use mirror split to separate a drive out. Remove disk and take home. What you are left with is a 2 disk mirror at work with a third drive elsewhere. You can periodically bring the disk back and resilver it into the three drive mirror.

[@hardcorepooka104]

Have you considered something like Syncthing for backing up the server to the Synology unit at your house?

[@poldelepel]

For small form factor problems you could contact optimum tech.

[@jabaniz]

Would you use a VM for encode/decode or just use like Handbrake docker instead?

[@eltonko]

Can't the workloads run in containers rather than vms?

[@aaaa-od4wn]

What are thoes yellow plastic GPU holders. never seen something like that for add in cards on shelf

[@mattsmechanicalssi5833]

Good question. Now I'm curious too!

[@bytezero3818]

i guess 3d printed, that's how most professional looking solutions are made, especaly if it is a problem noone else has

[@natterman2037]

I love how word things! You the man Steve👍❤✌

That being a server motherboard, If noise is not a problem you might want to take a look at Dynatron A18 or Dynatron T497 for a more silent option(and review both of them).

[@EmuAGR]

I got an X470D4U2 too, to make use of my old desktop R5 2600, the cooler clearance is definitely an issue if you are height-constrained as you are with that case. I also got an used Cryorig C7 before I saw Wendell's thread about the clearance, but at least it was cheap and I'm planning to use 2x 16GB UDIMM 2133 sticks. My choice was a 4u server case so I can stick an spare Wraith Prism to it, as it is taller and won't cover the slot. Or maybe just leave the Stealth. I also considered to stick an Asrock X470 Taichi Ultimate for its 10G port, but I wanted to test the IPMI giggles 140€ more spent on a tech whim... Wondering about installing UnRAID + ZFS. I'm also in my way to get a SAS DAS and cables to pair with an HP P822 SAS controller. I could've gone with just 6x SATA cables but then what's fun about using server-grade parts!

[@Sunlight91]

A few years ago I bought a 3TB HDD for back up, but still only use 140GB.

[@mannyc19]

Maybe look into Tape drive backup drives?

[@jasperrost78]

I have a very similar unraid server and when I was looking for HW I read that this mainboard handles 32gb ecc memory sticks fine. Might be worth considering if you have problems with cooler clearance.

[@ThePhoneix999]

How loud is that disk shelf? I'm thinking of buying one of them for my media server.

[@unijabnx2000]

FYI linus (ltt) has a truck load of 16TB drives to give out/away.

[@dvdv7777]

Plot twist: Linus has been hoarding those as payment to get a bunch of Gamers Nexus modmats.

[@eaghra1995]

not sure what you have for a cpu cooler in there now but i have the scythe big shuriken 3 in my sff build. works great.

[@johnpaulsen1849]

For your offsite copy backup. Check out the synology Active Backup software. It can do schedules and can compress and will use deduplication to reduce size of backups. I have run a Managed Service backup system for over 8 years. I always tell people at least 3 copies of you data. Production, backups(offsite) and replication. Also both raid and snapshots are not backup. Raid protects from simply hard drive failure. Snapshots can protect data, however the longer you keep a snapshot the more overhead io you have on the array and only protects you if the array is functional. Sounds like you have a good plan. Keep up the great work!

[@ElijahPerrin80]

I have been thinking of this alot lately actually but I am not a pro, I was thinking I would do a raid server with tape backup that is put in a safe deposit box at a bank when full, and a mirrored server at home so I have access to the data at home.

[@daa3417]

Is automated deletion something common in the video/content creation space? I just don’t think I could do it, I don’t fear automation but there are some functions I want to have sole control of and deletion is one of them. I can imagine all the ways it could go wrong, unlikely sure but I’m at the top of Murphy’s shortlist so I’m not tempting fate here.

[@CoreyKearney]

Steve, a Noctua u9s is spec'd for 2u, doesn't cover any ram slots and is rated @ 95w I have one cooling my OC'd 1600x

[@Wirenfeldt1990]

Twitter my dude..

[@shetho1]

Could you do a video on how you set up remote backup

[@Jorge2222]

What are your zvol stats? aka: zpool list -o name,size,alloc,free,cap,health,altroot VolName and zfs list -o name,used,logicalused,compressratio,referenced,logicalreferenced VolName ?

[@bBrain]

I've seen many derpy server racks by guys not really knowing what they are doing. I give that a 6 out of 10 derpy rank. Extra derpiness for using grandma's lamp table as a server support! :D

[@guywhoknows]

When doing the book on back up, there is so much options. So, you would work off say 1tb drive, you would have 2 tb live backup. So you have a copy and you have a time copy. So a daily back up and a weekly. You then take a daily every 2 days to a caddy drive 1tb which was then taken off site. A second was also made which was kept on site in a waterproof and fireproof safe. Zoned back ups were also on the group, say office was one, accounts the other, and a data (server) a copy was hidden on all computers from the group. So if the server went down or a workstation, you could pull data locally., If there was a issue there, the. You pull from the server, if there was a issue there you pulled from the backup on site. And if there was a issue then you pulled from off site. I remember having set up the backup code line, and a auto burn DVD. A dvd would be burnt at the end of day of the business data. The instruction was remove the disc and replace the disc. There was a spindle sat for a age. Then caddies came along due to size issues. It was a long time till most businesses got high speed to make use of networks. And on site meant two things. Back up online is always a pain due to the overhead and use, so it's a night run. I still use on and off-site backup even with hard data. I'm surprised you didn't have a 4u rather than that little box. Mine which I have now gone to two from 4. Holds 19 drives four are hot swaps. It fits a standard ATX and parts so building price is down to the drives themselves. It is full air filtered. Has a lock and has a fire rating and bomb blast rating. It's good to have two, and you keep one off, and turn it on to synchronise. So it has a copy. This is great if you want and requirements are 0 down time. And zero loss. Now.... NAS is not backup. Often used as live it is a high risk. Take a look at SAN. If you do SANS. Then it works differently. As you have your network, you have off site network and you certificate all devices. You then split data to each off site so that the data is useless and encrypted. You can keep half, which obviously reduce size for you. But you have to keep the keys else where. And you load balancing the entire system as a SAN and following backup. I'm.sure we wrote the directives on this which are detailed notes on "how to" and policy and procedures. VT is a bit different. I have five backups and no data is deleted until a set time from first point. So say, you have 100g drives recording. There would be 20 drives. They would be wiped before they are next used. Giving 20 days or events AS prime backup. The. Records were then loaded to a drive and not raided to another (sync) That's raw data. The edits were done and stored else where. Final was then done. We then kept the pre edit but not raw. Then The final was kept in reduction. Edit copy was kept as a temp. You just do a date scan and pull folders over or file when they get older. *** Back ups fail or are corrupt *** I've pulled from a back up a copy of corrupt data. The system can also write corrupt data. Always check validity of data. (Not file check) I've had files check, but they were damaged on playback. Buy everything twice... Is a fair rule. Do it again to make sure... Test all systems and fake break and see how life goes on.

[@user-nq3xi4wm2k]

Did you make those gpu holder or did you buy them, if you bought them where can I find them

[@unijabnx2000]

Did you discuss the use of proxmox over unraid?.. just wondering what made you want unraid over proxmox?

[@firsttyrell6484]

speaking of backups, Snazzy was recently featuring an LTO tape drive, looks like an overkill to me

[@aemonblackfyre4159]

First Tyrell looks like a perfect solution have to servers one backing up to the other and then you have a offline tape backup that you store in a different place only way you could lose all would be if something catastrophic happens while youre backing up to tape but well... theres no such thing as perfect safety

[@stevesadness8560]

hey Steve, seeing that your workplace is rapidly filling up, are you thinking about or even considering moving into a place similar to Jay's warehouse in California? Probably sounds premature since you haven't been there very long, but with staff increases and the growth of your channel, what would be the next logical step in your opinion? By the way awesome content on both channels.

[@bothellkenmore]

I think he touched on that on his Headset storage video a week ago. kzbin.info/www/bejne/qKPcd6OVhql5pK8 I'm pretty sure getting an adjoining suite and punching a doorway into it would be best case scenario for sure.

[@GamersNexus]

Great question. I'll take this for another video!

[@VaKo]

I do love watching desktop people build servers, its so different from how we actually build servers in the ops world.

[@tomtalk24]

No shit.

[@blackax]

For something you have to "Own" and maintain your plan will work out great

[@shetho1]

Have you got a video on setting up all of the server rack

[@Kraaketaer]

Might the Alpenföhn Black Ridge work for you as a replacement CPU cooler?

[@PATTON2189]

Hey Steve, What settings are you using for transcoding?

[@mbe102]

Hey, whats that Linksys stack you got going on there Steve?

[@THE-MAD-TECHIE]

Check out the Noctua NH-L9a-AM4 it is 114x92 vs your current cooler is 97x97 (so if you have 114 MM the long direction you should be fine.) you would need to check clearance Noctua also makes a 95x95mm cooler but it might still hang over the slots (the board might be a hair tighter fit than spec says it should be). Most AMD boards should have a specified 96x114mm keep-out zone (the C7 is advertized as fitting Keep out zone specs for Intel (which I believe is 95x95) yet is 1MM to large each direction , guess they figure most boards have a hair extra space) If your board meets the AMD spec the Noctua should work.

[@SerpentXTech]

Can I get a summons? need to go to the Auction house... I had an intriguing conversation with fellow Patreons regarding which NAS config I should go with for my personal home setup, I enjoy this type of content, thank you. Keep up the good work! P.s. It has been a minute since you uploaded and trail riding footage, I assume you just haven't had time. I got the Go Pro Hero 8 with Media Mod, which works great. For mountain biking you won't need the media mod, but the "Rollcage" would work for you.

[@unijabnx2000]

Have you talked to eposvox about how he compresses his video/files?

[@sonsofmatriarchy5552]

*Where's Wondell?*

[@LaughingOrange]

At his office or home.

[@Sbellins1109]

For security reason you can use a program that allow you do recover files deleted form the shared folder. Linus use this thing as he show in a video. In short: if someone cancel a file from an SMB share this files insted of benn deleted it will be moved to another folder, another disk or another server, as you prefer. In this way you can also avoid loosing data for human errors. I will be interested in this thing too but i cannot find a windows program to do this. If you find a way can you also make a video about this?

[@ArturoTabera]

notice the mandatory bike in the server room.

[@jasonkwong3207]

Great category choice!

[@evanparsons8757]

If it’s still the asrock x470d4u with the R5 3600 you can use the old fx 8350 air cooler and it will work perfectly. Running that with R7 3700x on mine for a home server

[@phillipramz8882]

What about EMP attack?

[@excitedbox5705]

I would love to see someone go crazy building a super backblaze storage pod with even more storage than the stock design. With your backup policy you have no need for that though.

[@nicks.youtube]

Any chance you played with VMs before getting it all setup for work? Curious how this setup would do running a Windows gaming VM and a Linux daily user, on top of light server duty

[@CheapBastard1988]

You're probably better off on the Level1Techs forum. If I remember correctly there were links to the applicable forum threads in the descriptions of the original video's. Unraid with an AMD GPU should be relatively straightforward for VM's though. Getting ZFS to work with Unraid was the most complicated part.

[@spewp]

Good thing you have that server rack, y'know for your desktop router & wifi.

[@zstation64]

That’s not a server rack. It’s barely a comms rack.

[@DurzoBlunts]

"Cooler" as in fan? Or heatsink and fan??? Because if just fan and since this is a server go with Delta they have a fan list longer than my Christmas list from 7. Should find one that meets your size requirements.

[@Rainbow__cookie]

Thanks KZbin for recommending this to me Now I'm going to subscribe

[@Rainbow__cookie]

@Wu Me I know gamer nexus just did not know this channel

[@shakingcow816]

Raid 1 is always nice, could look to buy a VXrail on Ebay or a VNX or VNXE (active passive). Can find them pretty cheap. To bad you cannot get in to SRDF or other products like this. I would love to see content on Big data arrays.

[@KannonRomano]

Remote site backup from unraid to unraid I would like to see that

[@slappopotamus1001]

I like the server stuff cuz Quite Frankly i can only watch GPU and CPU tests so many times in a row till it all runs together.

[@theatlastech8792]

I have only 1 question.. What do you use for your backups?

[@davifarhi1613]

for linux i would guess it is rclone or rsnapshot

[@mckidney1]

I like these videos :) Anyway. year later and I bet the Synology is still there :D Ramote Backup plans almost never materialize :D There is a local backup version of rdiff/rsync that allows you to bring a drive back and forth for semi-local. It is good for these as internet is slow and often in use.

[@packerbackermk1]

ZFS will use as much RAM as it can get it's hands on usually, but not always. So, if you install your RAM and wonder why it still seems like there is 90% utilization, it's because of how ZFS treats RAM.

[@dashtesla]

You should consider a tape drive, dump everything to tape and archive for long-term doesn't use electricity and you can keep everything you're actively using stored in the server. Really just stick a sas controller/hba into a computer and get yourself a drive, LTO 7 M8 or LTO 8 should be plenty good for your needs it also helps with any potential disaster you can keep more tapes offsite as well have a drive at home to move a lot of data back and forth.

[@gsydaz]

Wendell, the all knowing Gandalf of server building.

[@junkerzn7312]

ZFS sucks up all available memory. The memory probably isn't 'running low'. Though if you are running VM's then yes, the more memory you have the better off you are. 32GB sticks are fun :-) ... 64GB in a mini-itx two-dimm-slot motherboard or 128GB in a normal 4-slot motherboard. Normal or EUDIMMs. It just works (on AMD). I forgot whether you had set this up on Linux or Windows. If on Linux, though, you probably just aren't used to understanding how its memory management works. For dust... machines rooms aren't supposed to be dusty! I recommend buying an air purifier for your machine room (Winnix is a good brand). It can just be left turned on in low or sleep mode, since its a tiny room (if it has a PlasmaWave option, turn that off. The HEPA and carbon filter is plenty good enough). Better that then having hard drives have to filter the dust out. Those case dust filters don't really work very well in terms of the types of dust that HDDs are vulnerable to. I universally use external air purifiers for the room rather than dust filters in the cases, it's just easier to deal with. Another off-site option is to use a service, like BackBlaze. You are quickly going to find yourself with a multiplication of physical hard drives if you try to do the off-sites yourself, particularly if you want a setup that actually stays relevant and operational over years. For example, even if your off-site is RAIDed, if it just has a single logical copy of the data and it dies, resynchronizing from your masters over the internet can literally take forever... MONTHS or worse. Even with RAID, the chances of an off-site going poof and requiring a full resync are pretty high. For my personal storage, which isn't nearly as much as what you guys are trying to backup, I keep two copies at each off-site (two off-sites + the master location... 5 copies of the data total). Each copy is non-redundant and if it fails I replace the dead drive, reformat, and make a fresh copy from the other half. That way I am never faced with the prospect of having to re-synchronize over the internet. Since I'm not using redundancy for each individual copy putting them on the biggest possible drives keeps the drive count fairly low while still keeping the backups relevant and maintainable. For your purposes, a little redundancy for each copy is probably still appropriate, but don't go overboard. Copies x redundancy = too many physical drives and the law of statistics dictates that the failures will start to drive you crazy. -Matt

[@fie1329]

Ikea wood furniture is the best foundation for... almost everything!

[@elikirkwood4580]

personally i think that case does not justify its price when you can buy a Supermicro 4u chassis with as many hot swap drive bays as your Netapp DAS, and room for an eatx board and redundant PSUs for less than $200 on ebay

[@Propotus]

Do it with docker!

[@hiyan1]

"I bought this ram... this is... warm!"

[@lollerich]

Have you thought about having Patrick host some main channel videos? He seems like a good dude, and maybe that would allow you some free time?

[@AegisHyperon]

Bit odd to be using ZFS when it's not supported by unraid, guess you've got Wendell as an on-call support person.

[@jd_flick]

to automate the backups, you probably won't need to use cron, you can do that with systemd now. Site with info on service files: www.devdungeon.com/content/creating-systemd-service-files How to set up timers: wiki.archlinux.org/index.php/Systemd/Timers I'd suggest just setting up a custom script that uses rsync, making that executable and using the above info to automate it. Digital ocean has some good info here: www.digitalocean.com/community/tutorials/how-to-use-rsync-to-sync-local-and-remote-directories-on-a-vps

[@tristunalekzander5608]

What even is GN?

[@jasonhowe1697]

the storage capacity dictates the use of ram, 64GB most likely will not be enough for the caching value of the hard drives...

[@BeardedHardware]

Slacker!!

[@mattmilby]

I'm a Spectrum commercial technician in Raleigh. Many of my commercial customers have spectrum to back up their at&t internet or vice versa (depending on which one is faster). This is also true for Spectrum Business Class Fiber customers who have Spectrum Business (coax based) internet as a backup.

[@hazonku]

MOAR Dedudated WAM!!!!

[@ericbarry4656]

If there's ram in the system, ZFS is going to use it.

[@vcjester]

And then you have KZbin as your 4th backup... :P