I like the clean rack mount most of all. I just finished a project with two NUCs and agree on their handy utility! (And I'm glad Red Shirt Jeff has been pestering you lately... finally I can get some work done!)
Crazy you did this! Exactly what i did about 8 months ago. I had a dell R720, readynas, 2 switchs (with POE) AND BOY WAS MY POWER METER SPINNING. Got to the point where I didnt see the need to have such a power hungry server in the house and moved it to a data center where the space and bandwith was far cheaper than my costs to have it at home. But having this 3 node cluster was perfect for my needs and i could still do some low end development on the home lab if need be. Great video man looking forward to the next one!
I picked up an old R320 and it’s been interesting to play around with, but I completely agree the power demands are too high. I’m thinking of going the cluster route and just gutting the R320 case and mounting the NUCs on that. I already have the Dell rails and cable management arm.
Thanks for the video. I run a DN2820FYK Celeron now for more than 9 years. It was my entrypoint to Linux on a x64 platform. Evaluated numerous stuff on it. Learned about Docker and so on. NUCs are super reliable from my point of view and worth every penny over time.
Love this video, it's exactly into my hobby sphere. A NUC is great because of a small footprint. If you have a rack (and it's somewhere where a bit more noise doesn't matter), you'll get a lot more back for your buck with an actual server or 3. Typically, the dell/lenovo/hp machines are quite affordable, especially if you buy used/refurbished. I'm running 2 lenovo's with a 6-core I5, 64GB of RAM and 2 500GB SSD's each. For a homelab, anything "tiny mini micro" is great, and suits a K8s cluster quite well, but for me the real strength of a homelab comes from building experience in setup + teardown. From that perspective, getting 3 machines to form a Proxmox cluster would make a lot more sense. On this setup with virtualization, you could easily fit 4 or 5 clusters with a controlplane and 3 nodes per k8s cluster, and have a bit more room to play with. In any case - for anyone wanting to build more hosting / devops experience, I highly recommend finding a small machine, cram as much RAM into it as you can, and start setting stuff up. Right now I'm working on an "end to end terraform solution" where I setup 4 VM's on proxmox with cloud init on an ubuntu template, initiate k0s as a distro with 1 controlplane and 3 nodes, and then also automate all of the final details, like storage using nfs to my nas, metallb and ingress. Challenges always remain, but it's fun to perfect (and then toss out and start over with something else).
Relating to the remote KVM part: there are NUCs with full vPro functionality with remote KVM over VNC. They also got a web interface to power cycle etc. - and there are also command line tools to start them remotely
This is really good! I was just thinking of deploying a full functioning project locally because I don't have money to maintain cloud services fees while still in development! I think you just saved me my friend.
I'm not sure if I should be proud or afraid that I start to understand what you're talking about and envy your setup after joining devops 2 years ago...
Tim's K3s Ansible playbook really simplifies setting up an HA Kubernetes cluster. On Ubuntu, it might be faster to set up microk8s but after trying both, I'm like k3s slightly more because of lower idle CPU usage. And even with stateful apps (which is... uhm... pretty much everything?), you can still achieve almost-HA with a single replica in a HA cluster + replicated storage (e.g. OpenEBS Jiva or GlusterFS). Just need to set tolerations to something like 10 seconds. So if the node (that carries the pod) is down, 10 seconds later, a new pod will automatically spins up on another online node. That's HA enough for me.
I either ran into you or your doppelganger just now in downtown Houston. Couldn't put my finger on who I was reminded of before the convo ended but wanted to say thanks after my memory was sufficiently jogged. Have appreciated your work for some time now. ✌️
Tim! Great video. I’m going to have to explore this. I’ve been thinking of selling my 8 x Raspi 4B (8gb) cluster and doing something like this. By the way, for your last Cloud Init video, I put together a ansible playbook based on the video to automate the process. It was awesome, used it to make templates with a single click across five proxmox nodes. Thank you for all you’re doing! 👏
Those Thunderbolt ports will directly do 10gig+ Ethernet host to host (iPerf around 14). 3 hosts, 2 ports each so you can daisy chain. Takes some configuration, but it's a cheap way to get fast networking.
I am thinking to run 3 R86S but as a proxmox cluster the 2 10gig interfaces are really interesting, you can use couple of ways 1x for client, 1x for ceph network 2x for ceph ring network and on top of proxmox cluster you can of course run k8s nodes as well so it seems best of both worlds
That’s pretty awesome. As you can see from my profile picture. I use the little Lenovo Thinkcentre and I use them as proxmox cluster. The SFF desktop below them is my OPNSense box. As you mentioned, they’re not as power sipping as the pi but they’re still very low powered compared to full size desktop or rack mounted servers. I use docker for my containers just because I find it simpler but I’ve been meaning to start using Kubernetes. Again really cool video and it’s cool to see these micro sized computers serve a purpose and in my case given a second life. Keep up the good work and can’t wait for your next video. Cheers!
Is there a reason not to use Intel ME (vPro) and buy expensive KVM stuff? The Intel NUCs and many more mini PCs have already Intel ME build in, including the capability to remotely mount FD and CD Images.
I liked the Jeff Geerling "Red Shirt Jeff" cameo. I'm good with docker and have been using it for a few years but man, I gotta get off my butt and get into kubernetes.
This is a perfect setup I think. A bit expensive for me. I've bought NUC i5 8 Gen. It supports virtualization but not those fancy features. I think home server in NUC is an optimal solution.
Great video! I just had to comment about the cost though... holy crap that rack mount drawer is the same price as one of the NUC's! That's an insane cost for what is essentially a folded steel drawer.
Since those nucs have 2 Thunderbolt ports, would you please try and use them as network connections? If you connect them to each other you'll only need to modprobe the thunderbolt-net driver I think. (Be sure to use real Thunderbolt cables!) Maybe the speed will be really good! TB does 40Gbit/s after all, so if you get like 10G that would be amazing!
Got rid of my 7 boys Proxmox cluster. Now using only 3 AMD 4800U servers. Yes it's 15W each, yes it's 16CPUs each, yes it has 2.5Gb LAN. Only issue is that it has only 1 slot for NVME and no SSD disk. But it's a bomb and my proxmoxies say thank you daddy every time I see these 30 VMs dancing 🎉🎉🎉❤
I'm not sure if the NUC's are compatible, but there's lots of Intel processors with AMT that allow out of band management on the same interface. I do this on my OptiPlex desktops using the MeshCommander desktop application. This might be something to look into if you'd like to save the PiKVM for something else!
@@TechnoTim 1st: Another excellent video, thanks Tim. 2nd: I'm not 100% sure what I'd do with it, but I want it 3rd: Some Intel NUC models have vPro (at a steep price increase)
Planning to do this in the future too, but I think I'm going to wait for mainstream ARM to come a little further to get the power usage down even further. Though maybe I should do a cost/benefits analysis, maybe switching to NUCs now and then again later saves the hardware costs in power costs.
Assuming 20w each for the nucs, and 5w each for some future arm thing, and $0.20/kWh. That works out to a difference of 45w. 0.045*24*365 = 394.2kwh per year, at a cost increase of about $79 vs the theoretical arm setup. That's about 1 raspberrypi in a case with a PSU per year. Some cstate tuning, and maybe scheduling a shutdown would likely close the gap. I'm not sure which cpufreq governor Ubuntu server ships with by default, but switching from performance to ondemand would also save power. If Tim's text overlay of 9watts average is correct, it's closer to 140kwh savings, or about $28. These mobile CPUs in the usff form factors really are very energy efficient.
Thank you. I follow you and watch a lot of homelabbing content. This is the first time I've seen a Kubernetes node on three devices for homelab stuff. What I've gotten used to seeing is a single device with Proxmox installed and multiple VMs. What are the pros and cons of both approaches?
Just a question everybody seems to be let out (or I don't have the right understanding of HA) what about if a pod/service runs with a database and a node goes down? How does Kubernet know who is holding the latest database entry? Thanks 👍
I ran a 4x node proxmox cluster using older NUCs for…. years. Originally used a cheap NAS for iSCSI shared storage but upgraded that to a better NAS that supported multiple 1GB links. I’m have a vmware background so using VMs for everything was my normal. Had a couple of Pihole VMs, a file server, an OpenVPN server that eventually got replaced with Wireguard, NGINX proxy, and a server to do remote support. Idk why I wrote all that… Oh…. yeah I used older NUCs running old 3rd Gen i5s but loaded with 16gb RAM and 1TB SSDs. Ran that for like 5 years or so? I still have one of the nodes running a dockerized pihole instance alongside a raspberry pi 3 lol.
Not having ecc support is a deal breaker for me, the worst thing about it is that you may not know if there is a problem when memory corruption happened.
What do you think about using three iKOOLCORE R2 Max (either N100 or N305 variant) to do about the same (but a single master) for lightweight workloads plus a teinted extra node (thinking Minisforum MS-01) for more power hungry workloads? Could even make an operator to power them on/off according to demand, making efficiency even better (a single node of each teint would be needed 24/7)
3:21 LastPass? With all your servers and clusters I’m really surprised you don’t self host Bitwarden/Vaultwarden or similar self hosted password manager.
any plan for a followup on this? Interested in seeing how you go on with all the services you mentioned. How do you setup the storage and running traefik as ingress. Would be cool to see some homeservices for instance zigbee2mqtt with an dongle. You can add a tag to the one who has it connected, and let k3s move the container around.
This looks nice for a lab play with. For power saving, I find proxmox on PC better as I can run the same services on fewer machines. I can also put 10Gbe interfaces, hba cards, gpu’s for ML , and more in them.
Hello, I've just made a corosync and pacemaker cluster between two debian 12 machines and I want to install CasaOS on the cluster so the load is balanced between the machines. Is it possible to install CasaOS on the cluster and if it is, how?
Hello, I have a question about clusters, maybe will be useful in order to work in video edition with software like Davinci Resolve?Can you do an example of that in case it is possible? cause video edition in Davinci Fusion requieres a lot of machine and maybe if one person can use 2 or 3 cheap pcs instead of a monster pc maybe would be a cheap solution
3 x Lenovo Tiny m920q high availability, zfs replicated, proxmox cluster. They each have pice slot and I have Mellanox ConnectX-3 dual SFP+ installed in each. They each run 5.5 idle without the dual sfp+. With the dual SFP+ they idle around 11 watts. Mine each have 32gb ram, and currently 512GB NVME.. Will be upgrading the nvme soon enough. No ECC but I am okay with that.
Nice idea. BTW, it's much better to use new intel 12th gen low-power cpu like N100. A small pc just look like intel NUC with N100 inside, just cost about 100 US$, without mem and ssd.
That's exactly the path I'm going to take. I have 3 NUCs, a new one from Beelink and two really old ones from eBay which I'm looking to replace. The ones he linked are great but I can get an N100 from CyberGeek for half the price, more RAM and a TB SSD.
No it hasn’t. NUCS will be kicking around for years to come as the warranty will still be honored, much like what has happened to Intel Optane. Also, mini PCs are here to stay, so just swap in any mini pc and reasons remain the same.
I like the clean rack mount most of all. I just finished a project with two NUCs and agree on their handy utility! (And I'm glad Red Shirt Jeff has been pestering you lately... finally I can get some work done!)
Nice cameo
Thanks not Red Shirt Jeff!
Hi Jeff,I got you here!!
Yup the OG Jeff is here ❤
wow see Jeff here 🎉
Crazy you did this! Exactly what i did about 8 months ago. I had a dell R720, readynas, 2 switchs (with POE) AND BOY WAS MY POWER METER SPINNING. Got to the point where I didnt see the need to have such a power hungry server in the house and moved it to a data center where the space and bandwith was far cheaper than my costs to have it at home. But having this 3 node cluster was perfect for my needs and i could still do some low end development on the home lab if need be. Great video man looking forward to the next one!
I picked up an old R320 and it’s been interesting to play around with, but I completely agree the power demands are too high. I’m thinking of going the cluster route and just gutting the R320 case and mounting the NUCs on that. I already have the Dell rails and cable management arm.
My cluster is 3 NUCs for workers and 3 hp mini PCs for the control plane. Been using this setup for a while. Very stable
Literally how Chick-Fil-A stores run their edge clusters in-store. K8s on NUCs.
Thanks for the video. I run a DN2820FYK Celeron now for more than 9 years. It was my entrypoint to Linux on a x64 platform. Evaluated numerous stuff on it. Learned about Docker and so on. NUCs are super reliable from my point of view and worth every penny over time.
I'm so glad to see you getting into low(er) power boxes. Excited to watch this!
Love this video, it's exactly into my hobby sphere.
A NUC is great because of a small footprint. If you have a rack (and it's somewhere where a bit more noise doesn't matter), you'll get a lot more back for your buck with an actual server or 3. Typically, the dell/lenovo/hp machines are quite affordable, especially if you buy used/refurbished. I'm running 2 lenovo's with a 6-core I5, 64GB of RAM and 2 500GB SSD's each.
For a homelab, anything "tiny mini micro" is great, and suits a K8s cluster quite well, but for me the real strength of a homelab comes from building experience in setup + teardown.
From that perspective, getting 3 machines to form a Proxmox cluster would make a lot more sense. On this setup with virtualization, you could easily fit 4 or 5 clusters with a controlplane and 3 nodes per k8s cluster, and have a bit more room to play with.
In any case - for anyone wanting to build more hosting / devops experience, I highly recommend finding a small machine, cram as much RAM into it as you can, and start setting stuff up.
Right now I'm working on an "end to end terraform solution" where I setup 4 VM's on proxmox with cloud init on an ubuntu template, initiate k0s as a distro with 1 controlplane and 3 nodes, and then also automate all of the final details, like storage using nfs to my nas, metallb and ingress.
Challenges always remain, but it's fun to perfect (and then toss out and start over with something else).
Correct in all aspects, I just think he wanted to do something different and jump straight to the k8s side of things. Spot on though!
Relating to the remote KVM part: there are NUCs with full vPro functionality with remote KVM over VNC. They also got a web interface to power cycle etc. - and there are also command line tools to start them remotely
The problem is to find them - cannot find them anywhere on the internet (just simply nuc ...)
Drag race? Your NUC k3s cluster vs mine 💪🏼
Let's go! 🚀
I would love to see a livestream contest, judged by Tom from lawrence system.
The Mk1 hardware looks pretty nice. Glad to see you having lots of fun in the lab!
This is really good! I was just thinking of deploying a full functioning project locally because I don't have money to maintain cloud services fees while still in development! I think you just saved me my friend.
The unexpected Geerling cameo had me laughing out loud!
I'm not sure if I should be proud or afraid that I start to understand what you're talking about and envy your setup after joining devops 2 years ago...
I'm super glad I found your channel. This is the pure kind of nerdery I need in my life. Love it.
Tim's K3s Ansible playbook really simplifies setting up an HA Kubernetes cluster. On Ubuntu, it might be faster to set up microk8s but after trying both, I'm like k3s slightly more because of lower idle CPU usage.
And even with stateful apps (which is... uhm... pretty much everything?), you can still achieve almost-HA with a single replica in a HA cluster + replicated storage (e.g. OpenEBS Jiva or GlusterFS). Just need to set tolerations to something like 10 seconds. So if the node (that carries the pod) is down, 10 seconds later, a new pod will automatically spins up on another online node. That's HA enough for me.
I either ran into you or your doppelganger just now in downtown Houston. Couldn't put my finger on who I was reminded of before the convo ended but wanted to say thanks after my memory was sufficiently jogged. Have appreciated your work for some time now. ✌️
Hey, and thank you! Not me I am way up north but would appreciate that you came here to tell me!
@@TechnoTim haha well your twinner is here and in shorts.
Love the NUC line. Definitely a real tb port.
👍👍- phenomenal! Thanks Tim!
Tim! Great video. I’m going to have to explore this. I’ve been thinking of selling my 8 x Raspi 4B (8gb) cluster and doing something like this. By the way, for your last Cloud Init video, I put together a ansible playbook based on the video to automate the process. It was awesome, used it to make templates with a single click across five proxmox nodes. Thank you for all you’re doing! 👏
With 8GB Pis going for 220€, you can build more than a 3 node NUC cluster, lol
Found this channel researching NUCs...absolutely awesome content. Liked and Subbed! 🎉
Cool to see I’m not the only one clustering mini PC’s. 👍 those small form factor devices are quite powerful lately.
Nice work! I love those little 4"x4" boxes and have a few in my lab
I love the NUCs, they are awesome
Great video 👍 This makes me wanting to setup a kubernetes cluster myself 😊
Those Thunderbolt ports will directly do 10gig+ Ethernet host to host (iPerf around 14). 3 hosts, 2 ports each so you can daisy chain. Takes some configuration, but it's a cheap way to get fast networking.
Just wondering if I can use different spec of computer for each node.
I am thinking to run 3 R86S but as a proxmox cluster
the 2 10gig interfaces are really interesting, you can use couple of ways
1x for client, 1x for ceph network
2x for ceph ring network
and on top of proxmox cluster you can of course run k8s nodes as well so it seems best of both worlds
I love your videos even when it is not my game sometimes:) ❤ Keep on being awesome !
That’s pretty awesome. As you can see from my profile picture. I use the little Lenovo Thinkcentre and I use them as proxmox cluster. The SFF desktop below them is my OPNSense box. As you mentioned, they’re not as power sipping as the pi but they’re still very low powered compared to full size desktop or rack mounted servers. I use docker for my containers just because I find it simpler but I’ve been meaning to start using Kubernetes. Again really cool video and it’s cool to see these micro sized computers serve a purpose and in my case given a second life. Keep up the good work and can’t wait for your next video. Cheers!
What addons do you have in VS Code to help with Kubernetes?
"Red shirt" Jeff is a bit too much chaos.
Is there a reason not to use Intel ME (vPro) and buy expensive KVM stuff? The Intel NUCs and many more mini PCs have already Intel ME build in, including the capability to remotely mount FD and CD Images.
I liked the Jeff Geerling "Red Shirt Jeff" cameo.
I'm good with docker and have been using it for a few years but man, I gotta get off my butt and get into kubernetes.
such a professional video review,envy of your hardware!
Love all the low power pc content!
Great video, being learning a lot from you, thanks for your work. That storinator is using more power than my entire homelab :D
Heya, thx for the monitor! Cool vids as well!
Thanks man! Enjoy! That monitor makes an appearance in many of my videos!
This is a perfect setup I think. A bit expensive for me. I've bought NUC i5 8 Gen. It supports virtualization but not those fancy features. I think home server in NUC is an optimal solution.
Great video! I just had to comment about the cost though... holy crap that rack mount drawer is the same price as one of the NUC's! That's an insane cost for what is essentially a folded steel drawer.
Since those nucs have 2 Thunderbolt ports, would you please try and use them as network connections? If you connect them to each other you'll only need to modprobe the thunderbolt-net driver I think. (Be sure to use real Thunderbolt cables!) Maybe the speed will be really good! TB does 40Gbit/s after all, so if you get like 10G that would be amazing!
9:00 😂🤣😂🤣😂🤣😂🤣😂🤣THE BEST and HILARIOUS part.....!!!
TIM be aware of RED_SHIRT_Jeff .....🤫🤫🤫
Got rid of my 7 boys Proxmox cluster. Now using only 3 AMD 4800U servers. Yes it's 15W each, yes it's 16CPUs each, yes it has 2.5Gb LAN. Only issue is that it has only 1 slot for NVME and no SSD disk. But it's a bomb and my proxmoxies say thank you daddy every time I see these 30 VMs dancing 🎉🎉🎉❤
😂
I'm not sure if the NUC's are compatible, but there's lots of Intel processors with AMT that allow out of band management on the same interface. I do this on my OptiPlex desktops using the MeshCommander desktop application. This might be something to look into if you'd like to save the PiKVM for something else!
Thank you! When I bought the first NUC, I thought they came with AMT, which they didn't and the reason why I went PiKVM!
@@TechnoTim
1st: Another excellent video, thanks Tim.
2nd: I'm not 100% sure what I'd do with it, but I want it
3rd: Some Intel NUC models have vPro (at a steep price increase)
Planning to do this in the future too, but I think I'm going to wait for mainstream ARM to come a little further to get the power usage down even further. Though maybe I should do a cost/benefits analysis, maybe switching to NUCs now and then again later saves the hardware costs in power costs.
Assuming 20w each for the nucs, and 5w each for some future arm thing, and $0.20/kWh. That works out to a difference of 45w. 0.045*24*365 = 394.2kwh per year, at a cost increase of about $79 vs the theoretical arm setup. That's about 1 raspberrypi in a case with a PSU per year. Some cstate tuning, and maybe scheduling a shutdown would likely close the gap. I'm not sure which cpufreq governor Ubuntu server ships with by default, but switching from performance to ondemand would also save power.
If Tim's text overlay of 9watts average is correct, it's closer to 140kwh savings, or about $28. These mobile CPUs in the usff form factors really are very energy efficient.
Nice! Thank you
Thank you. I follow you and watch a lot of homelabbing content. This is the first time I've seen a Kubernetes node on three devices for homelab stuff. What I've gotten used to seeing is a single device with Proxmox installed and multiple VMs. What are the pros and cons of both approaches?
What cable is that supplying power & data to the SSD? It’s very interesting, I haven’t seen that before
Great video. If I can suggest wherever your reading from to be lower. Looks like your reading from the ceiling or something
Nice work !
At home I use 6 Lenovo Tiny as Proxmox Cluster to run a Docker Swarm, a K3S Cluster, ...
Just a question everybody seems to be let out (or I don't have the right understanding of HA) what about if a pod/service runs with a database and a node goes down? How does Kubernet know who is holding the latest database entry? Thanks 👍
Tim, dont mess with Red Shirt Jeff... :D
I ran a 4x node proxmox cluster using older NUCs for…. years. Originally used a cheap NAS for iSCSI shared storage but upgraded that to a better NAS that supported multiple 1GB links.
I’m have a vmware background so using VMs for everything was my normal. Had a couple of Pihole VMs, a file server, an OpenVPN server that eventually got replaced with Wireguard, NGINX proxy, and a server to do remote support.
Idk why I wrote all that…
Oh…. yeah I used older NUCs running old 3rd Gen i5s but loaded with 16gb RAM and 1TB SSDs. Ran that for like 5 years or so? I still have one of the nodes running a dockerized pihole instance alongside a raspberry pi 3 lol.
Not having ecc support is a deal breaker for me, the worst thing about it is that you may not know if there is a problem when memory corruption happened.
What do you think about using three iKOOLCORE R2 Max (either N100 or N305 variant) to do about the same (but a single master) for lightweight workloads plus a teinted extra node (thinking Minisforum MS-01) for more power hungry workloads?
Could even make an operator to power them on/off according to demand, making efficiency even better (a single node of each teint would be needed 24/7)
Great video, I have a question with the NUC's. What would you do if you want to add more storage as in external hard drives?
where did you get that large kvm?
Really looking forward to your next live on Thursday because i work Tuesday until 8 unless your going live later on in the day
Very useful video, if you have ansible running....
Hey Tim so where is you master node, not in one of your NUC, right?
No ECC? How come?
Hi Tim, not sure if you've done a video on it, but do you manage your Unifi network with either Terraform or Ansible as part of your home lab setup?
Could you run proxmox on each node and make each node a separate k3s cluster using VMs?
I have a question, would it be possible to create a VM that would otherwise use cores from several cluster devices?
RED SHIRT JEFF CAMEO
3:21 LastPass? With all your servers and clusters I’m really surprised you don’t self host Bitwarden/Vaultwarden or similar self hosted password manager.
What can you do with longhorn? Can you setup something like ceph over the 2.5G nics, or install 10G usb dongle nics for more speed?
any plan for a followup on this? Interested in seeing how you go on with all the services you mentioned. How do you setup the storage and running traefik as ingress. Would be cool to see some homeservices for instance zigbee2mqtt with an dongle. You can add a tag to the one who has it connected, and let k3s move the container around.
This looks nice for a lab play with. For power saving, I find proxmox on PC better as I can run the same services on fewer machines.
I can also put 10Gbe interfaces, hba cards, gpu’s for ML , and more in them.
Nice info, thanks for sharing :)
Tim. You need to stop making me have tech envy. I am trying to stop myself from buying a system76 meerkat or an Intel NUC for using at my desk.
I’m confused, what storage provider are you using?
Hello, I've just made a corosync and pacemaker cluster between two debian 12 machines and I want to install CasaOS on the cluster so the load is balanced between the machines. Is it possible to install CasaOS on the cluster and if it is, how?
09:00 chaos 😂❤🎉
Hope @jeffgeerling saw this
@TechnoTim how is this managing E / P cores with K3s/ K8s
How does kubernetes handle P and E cores in modern CPUs?
I've been keeping my homelab on 10th gen intel due to never getting a solid answer to this.
Hello, I have a question about clusters, maybe will be useful in order to work in video edition with software like Davinci Resolve?Can you do an example of that in case it is possible? cause video edition in Davinci Fusion requieres a lot of machine and maybe if one person can use 2 or 3 cheap pcs instead of a monster pc maybe would be a cheap solution
Great video Tim, thank you. I use K3sup to provision, but I'm very interested in your Ansible version.
Moving off VMs like this: we shall refer to this as LITERALISATION.
Intel vPro is probably integrated in those NUC so you could probably use AMT that for KVM 😊
Don't let red shirt Jeff near the server!
Did you look at charmed kubernetes options with JuJu? It even integrates quite nice with MAAS 😃
Vs RPi? You forgot to mention the architecture. It seems there are many containers that do not run on ARM processors.
Since NUCs have vPro technology, have you tried to use MeshCommander to access AMT/KVMoIP instead of using the pikvm?
Not all NUCs have VPRO / AMT. I found out the hardware, but it was a happy little accident because now I can use PiKVM to control everything!
deeeeez NUCS! haaaa GOTEEM!
I'm amused by your notion of wanting low power while I'm thinking about how sweet a R940 would be to have as a Truenas box. 🤪
My thing with nucs is the ram!
3 x Lenovo Tiny m920q high availability, zfs replicated, proxmox cluster. They each have pice slot and I have Mellanox ConnectX-3 dual SFP+ installed in each. They each run 5.5 idle without the dual sfp+. With the dual SFP+ they idle around 11 watts. Mine each have 32gb ram, and currently 512GB NVME.. Will be upgrading the nvme soon enough. No ECC but I am okay with that.
Gostaria que tivesse mais conectores SATA para fazer um NAS storage.
red shirt jeff cameo!!!! 😂
Does Intel NUCs support ECC Sodimms?
It does not
Let's be real the tidy cables never last long for us lol
Do they send worldwide?
Amazon should!
@@TechnoTim I mean the rack mount provider
should look into consul / nomad stack for something lighter weight than kubernetes
That is what I run at home on my NUCs, love it
Nice idea.
BTW, it's much better to use new intel 12th gen low-power cpu like N100.
A small pc just look like intel NUC with N100 inside, just cost about 100 US$, without mem and ssd.
That's exactly the path I'm going to take. I have 3 NUCs, a new one from Beelink and two really old ones from eBay which I'm looking to replace. The ones he linked are great but I can get an N100 from CyberGeek for half the price, more RAM and a TB SSD.
I see The Chaos isn't properly credited! 🤭
Are these devices POE powered?
Nope
No but I can see how you think they were, the power output in the UniFi screens was via their PDU and not a switch.
Who let Red Shirt Jeff out?
This seemed to have rapidly age like milk as Intel announced no more nucs. *But* there are other great options still!
No it hasn’t. NUCS will be kicking around for years to come as the warranty will still be honored, much like what has happened to Intel Optane. Also, mini PCs are here to stay, so just swap in any mini pc and reasons remain the same.
Cool but kind of expensive
Bahaha redshirt Jeff being a chaos monkey