Building a NAS for Joel Telling!
Vložit
- čas přidán 4. 01. 2024
- @3DPrintingNerd and I decided to combine forces to build the Ultimate Video Editing and Home Backup NAS. Check out the unboxing and 3D Prints for this build over on Joel's channel - • 3D Printing for HOMELA...
Grab yourself a Pint Glass or Coffee Tumbler at craftcomputing.store
Other videos you might be interested in...
What is RAID? - • What is RAID???
Setting up TrueNAS CORE - • TrueNAS CORE 12.0 Tuto...
Follow me on Mastodon @Craftcomputing@hostux.social
Support me on Patreon and get access to my exclusive Discord server. Chat with myself and the other hosts on Talking Heads all week long.
/ craftcomputing - Věda a technologie
One of these days I want Jeff to get someone in his show named Jeff just so we can hear “welcome to craft computing everyone and as always we’re Jeff”
I nominate Geerling
@@nnubbyup…. I agree
👀
Bonus points if they practice it to say it at the same time
Or, Jeff Geerling opens it with the usual line, and CC Jeff walks in after a second and is just like 'No, you're other Jeff', just to play off the joke that started at the 45 Drives event
Joel is such a positive-energy guy. He’s a great advocate for the 3D printing community. Glad to see this collab.
Great collaboration. Joel asking some great questions!
Dude that snapshot part in windows is insanely cool. Holy cow!
Can you do a video setting up the backup for TrueNas Scale? Also could you cover remote backups from one TrueNas Scale to another TrueNas Scale?
That would be super helpful. I’m about to set up a backup server myself.
I have one local backed up to another local (mirror copy) using replication, but I’d like to see a video about it too, to see if I’ve missed anything. Also I imagine a remote backup would be the same, just over VPN or a mesh network like TailScale.
Woo...love home tinkering/home lab/NAS stuff and 3D printing! Two of my hobbies in one video...awesome!
Keep em coming!!!!
This was such a pleasant conversation to watch. Some friends and myself are all in IT (two devs, an architect and two infrastructure engineers including me) and this really reminds me of those times when one of us discovers and explores some fascinating new technology or other that the others aren't really familiar with and starts explaining how it all works, which can go on for a long time because it's always "oooooh yeah, and here's another cool feature! Oh and wait, wait, you have to see this too, you're going to love this!" on repeat. I love people that are visible passionate about something, truly, it brightens my day and especially so when their passion happens to be in IT. They say variety is the spice of life but I argue that in fact passion, in all its many forms, is the true spice of life. Great video, thanks for the smiles. I truly hope your channel keeps growing, you deserve it for all the passion and knowledge that you share and for the effort you obviously put into your videos. I raise my glass of Balvenie to you, kind sir
Love ZFS snapshots. Hell, even if you get ransomware, "oh, you thought you deleted my backups? Joke's on you, I'll just roll back to yesterday"
Is that true and correct?
But that's only true if the ransomeware doesn't nuke your snapshots, or the hacker in some way has root privileges. Snapshots isn't backup if it's on the same system, then it's just replication. Even though it's on a different pool or storage media. Not bashing on snapshots, it's a killer feature. Just setup a proper backup system. Preferably with a pull configuration.
@@KS-wr8ub Yeah, I was mostly referring to if you had something on a PC or so that tried to nuke an SMB share or something, since end user devices are much more likely to get malware on them (due to user actions) than the NAS directly. Still possible for that to happen, but the likely scenario that most people would run into is some ransomware on their PC tried to nuke the files on the mounted SMB share.
@@TechGeek01 Yeah, I agree. But ransomware attacks on NAS systems are getting more common the more services people put on them and the more ports etc that people (stupidly) open. Like I said, ZFS is a killer feature. People just need to understand that it's not a magical savior from not having proper backups. 😅
You'd need root access, and possibly execute permissions on the NAS to alter snapshots. So long as your access is properly secure, there's little to no risk of the NAS nuking snapshots without your direct say so.
When you said "login with root" I was like, whaaaaa?! But I forgot that might still be an option on CORE. Been on SCALE for a while now.
Excellent video & solid collaboration
Interesting choice to go with core versus scale...
Great stuff, more collaborations please and thank you.
This is super helpful! Thanks for sharing
Great to see someone go through the features nice and slowly.
It's kind of crazy to think that NAS units started out as just USB HDD enclosers with a LAN port added so you also had the option to SMB to your files off multiple machines, a great alternative to having to have a full blown home/small business server, then we got two and even four drive models, and they made such an impact that most people stopped building home servers and Windows Home Server was discontinued. And now we've got to the point where pretty much all the "NAS" units you see are actually full blown mini servers running applications and VMs yet we still call them a NAS xD
Nice video! Great intro to truenas as well. I like the format
I really loved this episode! I learned a lot how you guys use ZFS!
Been watching you both for a while, such a cool thing to see!
Seriously. Quality collaboration.
It's good to know that the cheapest HL15 config can handle TrueNAS [Scale] and 10G assuming you don't do anything else with it, as I'm leaning toward having a separate Proxmox box (or three of them for a cluster if I ever get the funds) to host everything else.
Could we get a breakdown of the settings used for Time Machine backups? I've given this a swing prior with OMV and its been spotty at best :(
Love it, i learn a lot !
I've expected to see in this video building hypervisor with virtual NAS and virtual rendering workstation, but it was just another video about cooking simple TrueNAS instance.
And why do you use rsync instead zfs replication task? Does rsync has ability to non-blocking access to file, which is in heavy editing process?
Great clip....
The "Previous Version" is an interface you get in Windows, and typically is called into play with Volume Shadow Copy. Here you are using it to access ZFS snapshots.
I don't use Windows anywhere. I am almost exclusively Linux, and I have one remote Mac Mini M1 that I remote connect to via RDP thanks to a combination of NuoRDS Remote Desktop for Mac and Guacamole. So how would one restore a file from a snapshot if you were using a Linux desktop with a file manager like Nemo or on a Mac with Finder?
I want to know the same thing!
Wow, I glanced at my youtube feed and saw Jeff on three different channels. Way to go!
December was QUITE busy :-D
cool build!
Man now this is a great video!! I will probably eventually go with True NAS as right now I have a NAS box specifically a QNAP the issue is that it only has 4GB of RAM which affects the applications running on the box, but as of right now is not affecting transfer speeds, love this video!!
Can you post a parts list please
Crazy I also worked for Mary Bridge Childrens Hospital for a while about 8 years ago doing IT.
Great video
Cool stuff
Awesome video!
That hand gesture in the thumbnail means something worse in the UK 😅
What does it mean lol? I’m pretty sure he just meant a peace sign.
@@WilliamCollier it's a two finger salute. Same as flipping someone off.
"If the palm of the hand is facing inward toward the signer (i.e. the back of the hand faces the observer), the sign can mean:
An insulting gesture in Australia, New Zealand, the Republic of Ireland, South Africa, and the United Kingdom.
@@MenaceInc I'm from South Africa, and I've never heard of it being considered something negative locally.
@@boneappletee6416 seems like usage in the Commonwealth has dropped off significantly in the last 40 years or so ¯\_(ツ)_/¯
I've currently got truenas running on my unraid, I wish I had the space and such to make it it's own standalone machine then I'd be more inclined to mess with it and set it up for home storage compared to the little storage I currently have it as.
No beer in the desk? Inconceivable!
Great video!!! Do you have a written instruction of this?
great videos both of you! Unraid has ZFS, is there a difference or are they the same? thank you
Automatation: Automatically replicating all your mistakes.
great video and love the work. I have one question, is there a specific reason you decided to use Truenas Core vs Truenas Scale?
I hadn't realized about the time machine flag. I know what I'm doing over lunch.
Awesome video! Long time viewer of @CraftComputing from when Jeff was building the rack and PCs in his closet in the early days of the channel and have learnt a lot about homelabs from his videos.
One of my hesitancies with zFS and systems like the ones shown in this video is "my understanding" that ALL drives used need to be identical. So, it is great that the software allows two disk failures but what if the disks used in the system are no longer manufactured / available? For example, if I start with let's say 6 2TB drives and then, when I have filled my NAS with about 4.5TB of data, one of the 2TB drive fails few months / years down the road. By now, however, the smallest capacity drive available from that manufacturer are 4TB drives. Does that mean I would have to replace all 2TB drives with 4TB drives in the NAS? And in the process, temporarily copy all the 4.5TB of data off the 2TB drives in the NAS to say an external drive, replace the 2TB drives with 4TB drives, have TrueNAS rebuild the "pool" and then copy all the 4.5TB of data back to the pool?
Again - like I said above, this is "my understanding" of how zFS / TrueNAS pools might work in that they require identical drives. And it is possible that there might be some gaps (misunderstandings?) in "my understanding" of strategies for long term maintenance / replacement of failed drives. Just posting here in hopes of learning / filling any gaps in my understanding. Thanks to anyone in the community for any feedback.
And thanks to Jeff @CraftComputing for this video showcasing a real world application/implementation of great OTS hardware + Open source sofware. Keep up the great work 👍
As someone that's had many ZFS pools over the years, let me answer some basic things.
"All drives need to be the same". This is absolutely not true. That's best practice, but I have mixed pools and they work fine. You'll get the performance of the slowest drive as it has to sync up while running.
As for expansion, you can expand storage in several ways, but the most common is adding a new vdev. If you have for example 2x2TB in a mirror, and want to add 4 more logical TB, you can add 2x4TB as a mirror, and you'll have 6 total logical (12 physical) TB of storage in 2 vdevs, but 1 pool. Other option is to replace all disks in a vdev with a bigger capacity (one at a time of course) and once the drives are all physically replaced and the pool is rebuilt, you can set ZFS to expand the vdev, and it will make the rest of the space available. It will take the smallest drive in a vdev as the size for all drives in a vdev. If you did a RaidZ2 (Raid6) with 2 2TB, 2 3TB, and 1 10TB, it would be considered as 5 2TB drives. If you later replace the 2 2TB, and expand, it would become 5 3TB drives. It's way more expandable than people give it credit for. Just know that if any vdev is fully dead, the entire pool becomes dead, so be careful.
@CraftComputing Question... Core vs. Scale in this case. Scale still not quite ready for a production environment, or was there another reason?
Can't go wrong with FreeBSD
Joel just have more BSD in the house
Is the a gdrive sync client integrated in TrueNas?
love the video man i got a unraid build with 10gig networking and this makes me want to build a true nas build now with 10gig networking to compare the 2 lol
If you are saturating the 10gig with unraid, truenas won't do any difference. I run truenas myself so no bias against unraid.
When he was listing his grievances with Unraid it just sounds like he has no idea he could greatly improve his workflow with Unraid. A minor annoyance of mine but his ignorance makes it seem like Unraid is not a good option. It sounds more like he was cheap and built a shit Unraid box.
So cool, I need to configure my snapshots better
How many watts does it pull when running a typical workload or dling the drives? Looking for something like this but electricity is quite expensive here in germany (0,30 Eurocents at the moment / is about 33 US cents at the moment)
Lol Kronenburg Jeff, on a Rick and Morty Episode fixing Ricks interdimensional NAS. So Rick can get some file containing a project Summer accidentally deleted ft. Kronenburg Kraft Computing. Im down for it.
I would run snapshots much more often than the archive copy (and also enable them on the archive drive)
Nice video. First time I ever noticed an edit. The used space went from 50GB to 5TB right after the Time Machine talk. I guess Joe backed up some files between calls?
We set up Quotas on that share during our conversation 😊
You would be surprised how frequently you will lose several drives in the same month. Since most of the time, all your drivers come in at the same time from the same manufacture from the same batch of production, so they all tend to die at around the same time.
Jeff, when you set up the pools on the server I noticed 2 things.
First the system dataset was on the HDD pool, likely the first set up. Would you move that to the SSD pool.
Secondly I noticed that you established multiple datasets in the top level set - I have done the same but have read multiple times that setting up a root dataset, say zroot, to give mnt/rust/zroot is preferable because if you then use zfs replication you can replicate zroot on one system to zroot on another whereas TrueNAS cracks one if you try to replicate rust to rust rather than rust/zpool to rust/zpool. You have to perform rust/dataset1 to rust/dataset1, rust/dataset2 to rust/dataset2 etc on a backup server.
Be interested on your thoughts. Thanks as always, so to speak.
5:00 By the way Fun Fact :- "My Favourite Colour is both Green and Blue....!!!😉"
Just wondering why Core version and not the Scale one?
Get it off the floor !
Dust management 👍
@CraftComputing or anyone who has a similar setup, can you provide the rsync script, switches, flags used in this particular setup from 23:50 timestamp?
I am looking for the same thing!
True nas core is the best/fastest more stable solution for storage needs.
What laptop are you using in the video?
GPD Win Max
@@Conor42203 thank you
@@coppertop27 specifically the Win Max 2 2023
im a lil jelly of that symmetrical 1G fiber
Same
Same
One benefit of living in the Seattle area. We get access to gig fiber but have to pay very high housing costs for the privilege.
Doesn't Unraid support ZFS now? Would it be possible to do something similar via Unraid's ZFS support?
Yes it does now
yes and its better then this because it can live inside parity protected space and outside in same time so you still can upgrade 1 hdd at a time...
@@IceSerbia Thanks. Been running an Unraid server in one form or another for over a decade now, but ZFS is totally new to me. Definitely see the advantages though.
Is that a strange Parts sticker? 5:36
So how many snapshots you configured to store? 24*14=336? And how fast it will eat all space during working on a project?
It won't eat up space because snapshots are "thin" copies. The filesystem only stores one physical copy. It doesn't make a physical second copy until you change the data.
@@eDoc2020 Yep. And during video editing lots of data change :)
@@Prime0pt There shouldn't be much data changing during video editing. The vast majority of space is taken by video files. The source files don't change and the output files only change when you render out the final project. Unless you are making tons of extra-high-quality test renders you won't fill up the space.
Why a crontab rsync instead of an Rsync Task?
I just named my zpool I setup yesterday to Rust also. 😆
What laptop is that?
Isn't technology great? You can describe what it can do for you in a matter of minutes and then take days to implement it because it never works out that way at first. The art of melding hardware and software.😊
I just love the size of his laptop compared to him and the server
5:10 "We blue ourselves."
Pretty surprised you used Truenas Core. Why did you pick Core over Scale?
Why Scale instead of Core?
A video where Jeff mentions RAID and doesn't mention it's not a backup???
I said that on Joel's video :-)
I had an 8 drive raid10 array on my homelab veeam server.
2 drives died at the same time.
Guess which drives died and what happened 🤣
(I did not play the lottery that weekend)
Please still have some sort of off-site backup, snapshots etc are great but they don't survive a fire
Wait now, that "rsync -av /source /destination" will only give him redundancy in the archive dataset of the last hours "snapshot" right?. Meaning, if his Sonic pool for some reason took a dump, he'll loose any snapshot redundancy. Why not setup this cron job to do deduplication and pruning on the archive allowing him to go back in time also in the archive. Might not be within his needs of course, but I figure why not since you're already doing the sync. Wouldn't have to take up that much more data. ☺
(with respect for us not seeing the whole rsync command)
Also, please advise Joel to don't use Google Drive for backups. Use something proper like Backblaze or Amazon Glacier. 😅
He might already have that. He didn't say if he was doing snapshots on the archive also. I wouldn't do deduplication though - that's when you start needing crazy amounts of memory for ZFS. And for most of what most people are doing, the block level COW ZFS does for snapshots is sufficient. The main reason to do dedupe is if you have stuff like lots of VMs with the same content but ZFS can't necessarily know that.
What I'd do (what I do, actually) for backups is setup another TrueNAS box at some offsite location and use TrueNAS's replication abilities. Replicate once per day at some odd hour. Works pretty well and nicely integrated into the UI. Also, since it uses filesystem level knowledge, it only needs to transfer changed blocks and it can figure out what blocks those are quite quickly. That's an area I've seen rsync fall over on - if you have lots of tiny files, an rsync can take forever, even though the data really isn't that big.
@@ccoder4953 I would still consider dedup here. If Joel changes the name of a file on "Sonic" it will currently get copied as another physical copy on archive.
@@eDoc2020 But what's the cost of a few extra GB of storage vs the extra GB of RAM needed to make dedup work properly? Per the TrueNAS documentation, a commonly cited number is 5GB per 1TB of HDD. That's the issue. Unless I have a use case where I KNOW dedup is going to be used extensively, I'd rather have that RAM for cache than storing dedup tables that will rarely ever hit collisions and just eat the cost of having duplicate copies on disk in the rare cases that I actually do have duplicated files (well, blocks actually). See, for dedup to work right and not be a massive performance bottleneck, you have to have tables in RAM or very fast flash with hashes of every block on the disk. If you're hitting collisions all the time, great, you saved a bunch of disk space. Otherwise, that RAM is really wasted. Does that make much sense for saving a little space during the odd times that a rename happens AND it happens to get archived? Doesn't seem like it to me. I'd rather put the money I would have spent on dedup RAM on bigger disks and more cache RAM - I'll probably net far more storage than dedup would have given me AND a faster NAS.
@@ccoder4953 The collisions would happen _any_ time you rename a file because it's always archiving. If remote editors their own local copy and then copy it back the space will add up, too.
Regarding the hardware needed to cache the hashes, look at the actual numbers. Using that 5GB/TB rule you'd only need 800GB for the entire 160TB rust pool which can easily be handled by a single consumer SSD. It doesn't need extreme performance, nobody is running an enterprise database on the rust pool.
On the flash pool "working directory" dedup would have less benefit because the data is only getting stored there temporarily and because performance matters more there.
Given the nature of this particular use case the performance hit of dedup could also be greatly reduced by using a larger block size. Choosing 64k instead of 8k would make a _huge_ decrease in requirements with little downside.
Hmmm…what do you call the SSD array consisting of 5 drives for a guy who lives “blue” and replaced OEM parts with 3D printed parts in High Five Blue PLA?
High5 !
Hi Joel, High Five
2 is not for two locations. 3-2-1 3 - Three copies, 2 - Two Formats, 1 - One copy off-site.
The snapshots should be on another system alltogether if they are considered backups.
(Preferably offsite).
It definitely does not follow the 3-2-1 principal.
"What is the 3-2-1 Rule? · There should be 3 copies of data · On 2 different media · With 1 copy being off-site"
We got the 3 and the 2, and I mentioned ideally we'd get his data offsite as well.
This kids is what is known as overkill.
which Laptop is that?
GPD Win Max 2 (6800U)
amzn.to/3tHN0Hf
Dabu dee, dabu blue dye
why truenas Core and not Scale (please help me understand)
Core is a much more stable version than Scale. Also, you can use your memory to the max for caching, unlike Scale (because ZFS works differently in Linux)
understood
Thank you @@franciscooteiza 👍
Well guess what 4th XD
FYI we have nothing but issues with Zoom speciffically for audio and video.
Edit, medium business.
It was an issue with OBS during the recording. I use Zoom weekly for my Podcast cohosts, and it's been nearly flawless for that.
I really need your help and not sure where else I could turn. My wife just got new offices for her business. And I set up the entire network. I have a rise in nine running the domain controller on windows server, And two client pcs. My issue is folder migration. I have ten terabytes of 2.5 in sata drives inside of an icydock drive bay in the server. They are slow as all hell. 20mins per GB copy speeds. It takes about fifteen minutes just to sign into a user account because of folder migration being slow. I know I need to get faster hard drives and a separate Nas, But only have a four hundred dollar budget. Can a raspberry pie with a u s b drive to be faster on the network? Even chat g p t doesn't have an answer for creating in affordable nas solution that as least has 2.5gb speeds. Please help me!!!
GPD Win Max 2 2023 spotted.
Early squad!
damn it, 2nd
Miss joke… “the damm pen is blue”
So … what exactly was built in this video ?
You just typed a freaking password in a login screen …
I think it might be enough for an HL15 promo.
So, you're moving about 100TB of existing storage to a device that will have only about 130TB of storage? That doesn't give you much room to expand. Why not use larger capacity spinning hard drives, to get a larger main archive pool?
Soooo... Why Core and not Scale?!
because it's BETTER! Scale is buggy as hell!
So we are no longer supporting your shenanigans anymore ~Jeff?
on the video thumbnail was very clear he want 2 not 1
Or he didn't like the camera
1st ;-)
Building a NAS for Joel? Hmmm, very telling...
are you gay?
Now just add firearms Jeff and you have the quadfecta of awesome hobbies! 😂
I'm not a big fan of you calling SSDs and HDDs on the same system "part of a good 3-2-1 backup strategy".
If you get a cryptolocker on one of your systems and that system is logged into both of your pools (as shown in your demo), it will jump to those network drives and you lose access to that data.
I'm also pretty sure that if you're only logged into your SSD pool and a cryptolocker locks down all your files on your SSD, when the next rsync/snapshot tasks kick off it will see those files as "changed" and update both your snapshot AND your copy on your spinning rust, introducing the cryptolocker to your HDD pool.
That system is copy of your data. Yes, the bits physically reside on two different sets of disks on two different mediums, but they're fundamentally the SAME data on one system, reducing your number of redundancies from a 3-2-1 backup strategy to a simple 2-1 backup strategy.
With snapshots, it's literally two clicks to revert every file back to a known good version if cryptolocked.
Both pools have Snapshots enabled, so even in the event the SSD pool is cryptolocked and that data moves to archive, you can select the last known good snapshot and revert the entire file system.
This is still TWO copies of the data, with snapshots ensuring against accidental deletion or malicious destruction. I know everyone would LOVE to deploy a by-the-books 3-2-1 backup solution, and that can certainly be done here for only 3x the cost. For a one-man organization like Joel's and Mine, this is more than adequate given the risks our data actually face.
@@CraftComputing In my days working at a MSP we had instances of cryptolockers that sat dormant for 3-6 months prior to activating.
Meaning that you revert to 1-2 days ago to a so-called non-cryptolocked version of the dataset and in 1-2 days the cryptolocker is back.
There is no way that an individual will be able to figure out how long the cryptolocker has been sitting dormant AND no one wants to lose 3-6 months worth of work.
The scammers are getting super crafty and this data storage setup is dangerous.
Popular CZcamsrs often need more than 15 drives of storage. Which is why the HL15 is overpriced.
why core though?
With the rsync push, in the event that the production dataset contains a different file (i.e. IMG_0001 from an iphone) with the same name as a file that already exists (i.e. IMG_0001 from an ipad) on the archive dataset, wouldn't that overwrite and effectively delete the file that existed there ?
Only if the file with the same name is also in the exact same path.
@@danielfisher1515 definitely not impossible depending on how he manages directories. It’s definitely something to be mindful of.
Both Joel and I use unique project names, starting with either dates or video numbers to prevent duplicate project folders. Just avoid naming your root folder 'New New Whonix' and you should be good.