Raspberry Pi Cluster Ep 4 - Minecraft, Pi-hole, Grafana + MORE!

[@CaptainCobvious]

Mate, your channel is obscure which in itself is a crime. Keep up the great work, man

[@AlwaysBolttheBird]

To be fair he seems to do very niche and more advanced than the average person does with a PI. I'm happy his channel seems to be growing so we can get more videos.

[@OmikronClient]

Wait guy i've seen you in some place 🤔

[@garylifosjoe]

Every thing you talk about or I have seen are bits and pieces for me.That I have done repairs on IBM servers,Cisco routers,servers and data switchers and I have worked also Dasan Zhone.Most of the time they give you troubleshooting software.Dasan Zhone was different.It was more like boards that can work individually or put more than 1 work together.Cisco has racks and you have a master and a slave that controle die other boards.I have also work on Sycamore troubleshooting with provided software.

[@draco5991rep]

You are the first person I find to actually explain in detail how to use a cluster and not only what you could use a cluster for. It is very easy to follow you (even so I don't know nothing about the drupal and Prometheus and all these crazy applications). The video is very well structured and is over all very helpful. Keep up the good work.

[@hillandrewdavid]

I've been working in IT for 20 years and I recently found your channel...totally hooked!! Thanks so much for your concise explanations, and mixing in some fun too :-)

[@josephturi]

43:28 You suppose to dig straight down in Minecraft. Coffin music starts playing. Brilliant.

[@PhattyMo]

And don't forget to hug a creeper,they need friends too.

[@first-thoughtgiver-of-will2456]

Warning: be careful with the Pihole Kubernetes deployment. If not done carefully it will break your Kubernetes DNS and can cause your cluster to become unstable. I lost a Ceph cluster by zipping through the docs and rushing a Pihole deployment after watching this. Personally, I recommend keeping your external network and internal network separate with respect to Kubernetes and LAN/WAN and use NodePort and Ingress where appropriate for data pipelines and servers-- that's it. The loss of my cluster was due to compounding errors with multiple chances for recovery and is completely my fault I am merely sharing my best practices. Thanks Jeff for the videos I really like this idea and am impressed with your card layout, power management and networking decisions therein. This would make an awesome dedicated Spark resource negotiation cluster as well. Tell Jason I said hi.

[@mojo_jojo]

Thank you for making a video about my pihole helm chart! :)

[@Samuel.Mwangi]

Fact Check: I have learnt more by watching that most getting started with K8s tutorials. Fact check: Jeff had way too much fun shooting and editing this video it shows. Too bad I can't like it twice

[@RonnyNussbaum]

What are the "getting started with K8s tutorials"? Are you referring to these videos or to something else?

[@Samuel.Mwangi]

Other videos on the subject.

[@RonnyNussbaum]

@@Samuel.Mwangi Thanks. Any specific recommendations that you have?

[@Samuel.Mwangi]

@Ronny N I made the comment 8 months at a time when I was in the process of learning K8S. Besides this video, Jeff has done an amazing series titled Kubernetes 101 series that combined with his book were incredible. Not sure if this answers your question on recommendations

[@RonnyNussbaum]

@@Samuel.Mwangi It answers it. Thank you. Yeah... I'm viewing that video series now, so thanks for your reply and suggestion!

[@vcokltfre]

Love the silicon valley reference lol, great video!

[@deepinsource]

OMG, just notice you are the guy 'geerlingguy', the majority of my ansible playbooks are from you Great work done!

[@HardestManInTarot]

My favourite Steve Buscemi/Raspberry Pi channel by far.

[@gaganrajput5836]

I don't even own a rasberry pi, but learned so many things from this video. Thanks Jeff!

[@gerrygreen3601]

I've watched your videos for a couple of months after acquiring a bunch of raspberry pi's to setup a kubernetes cluster. I do kubernetes in my day job and I have various home projects going including a cluster launched with vagrant and virtualbox VMs. I came back to youtube tonight because I thought I saw prometheus mentioned in one of your videos. I just finished a cluster using kubadm to launch a full k8s cluster on RPI as opposed to k3s. Then I used helm to install prometheus and hit the dreaded "exec format error" - the node exporters seemed okay but the rest of prometheus did not. I found it funny at the 10 minute point in the video that was exactly what I had encountered an hour ago. Thank you so much for this - great stuff.

[@cultureoverall3012]

excuse my french dude, but fuck your videos are so good. Your pacing is on point, your solutions themselves are good, the content is VALUABLE. like every time I pull up or come across a video from you it's distinctly good.

[@gabyabed]

Even though I don't do anything with Raspberry Pi, your kubernetes explanations are gold and had me engaged and learning. Thank you!

[@welovedotnet]

Awesome videos Jeff! With docker now releasing multiplatform builds through buildx a lot more easily, plus the new Macbook's using ARM processors, hopefully exec errors will become a thing of the past!

[@dwaynehart1974]

Another great educational experience Jeff. I'm looking forward to the next episode.

[@SlyEcho]

This is like the first tutorial I've seen that shows how kubernetes really works.

[@hoeurd]

You are an insanely good speaker. You should think about making podcasts. Thanks for the great content!

[@DJBioMusic]

Dude, I look forward to your videos so much. Especially the r pi cluster content. Keep it up!

[@PavelRozentsvet]

Thank you for the excellent content. I learned a lot from the Turing Cluster project as well as from The Mighty Drumble. Good Job!

[@TheMadMagician87]

This is a great series Jeff, thanks for all the effort you put into it!

[@mamba0815a]

Hi Jeff, many thanks for the excellent Turing Pi videos. I have a Turing Pi myself and, among other, both K3s and Grafana/Prometheus are running on it. One thing I noticed after two months of playing with it, is that prometheus-k8s is growing in memory demand over time. So much so it would crash on the Pi3 worker due to it little 1 GB of RAM (read this from the pod logs). As a result Grafana would wind up with no data to present. Although nothing else is runing on the prometheus' worker node and the fact I have maximzed the available RAM (i.e. reduced GPU memory) , 1 GB is simply not enough. I then tossed in a P4 worker with 4 GB of RAM and sure enough, K3S moved the prometheus pod over there and the cluster is stable ever since. I just wanted to mention that to you and your audience even if it may not be a surprise that the Pi3 has RAM restrictions. Having written that, I would like to relay a big thank you to you. Please continue your good work, I can't wait for your next postings. :-) Regards, Jay (Germany)

Tip: Using "su -" instead of just "su" brings all the environment variables of the root user and automatically changes the working directory to /root.

[@TerenceKearns]

I love being from the future, if I was watching this when it came out, I would have had to wait weeks for each nail-biting episode. Now I get to binge watch :) I still have no idea why I am watching this.

[@heitortremor]

This is a very well made video. Clean and informative. I didn't do any of the things with my lack of a cluster to do things with, but I do want to set something similar up in the future so this stuff is very fun to watch!

[@mayurchavhan8590]

Been waiting for this and it is rewarding for sure. Thanks Jeff for sharing this wonderful knowledge.

[@leep278]

Thanks for these videos! It's been fun walking through this episode.

[@gngn2973]

Absolutely fantastic video. Learned more about kube then I have from actual learn kube videos.

[@klintkrossa6885]

If I remember in the 80's - 90's there was a cluster computer being used to join two radio telescope to a single image. I wonder if the files would be available?

[@thisguy9647]

34:03 I like how he explains Minecraft as if there are people who are unfamiliar with it.

[@romeroadames]

Thanks!

[@AttosLins]

I dunno why or how I've been watching this whole series while not subscribed. Please don't send me to jail for this crime against humanity. Subscribe button smashed to bits.

[@ahmedexmor]

My friend you're really great at explaining. Nice work!

[@Whalebarf79]

‘Build your own’ I like that phrase. It’s like getting smacked with a glove in the face. Friendly competition is very healthy.

[@technoe02]

Dude, the Venn diagram broke me haha.

[@JeffGeerling]

Very lonely in the middle 😭

[@Jhare93]

i am looking into using pi for the first time i love to do diy projects and the way you ex[lain things very well makes me understand this more than the little but of knowledge i already had. I have heard the term retro pi thrown around alot as well and always wondered what that was.

[@settheshallow8913]

"Spaces are better than tabs" I WILL FIGHT YOU

[@porovaara]

He's not wrong though.

[@MikoKnight]

@@porovaara care to explain why? I don't understand why that would be the case. Tab just seems like a standard to me because it's always 1 unit instead of multiple.

[@lmaoroflcopter]

@@MikoKnight it's not as portable. Tabs can vary in size depending on the ide/editor you're using. Resulting in differences in how code looks across different editors. Softtabs/spaces are the smallest standard unit kzbin.info/www/bejne/iaTSgHpsj8qsi6s I don't know anyone who actually types spaces, I just use tabs, but configure the editor to type spaces according to the style guide of the code I'm developing.

[@a.mathis9454]

porovaara yes he is, very. Try debugging 15+ pages of code (especially printed out code), easier with tabs to know which lines go to which loop(etc).

[@slycordinator]

@@lmaoroflcopter And by the same token, spaces are less portable. You'll sometimes have to open code/scripts from a terminal editor that's pretty limited or just in some text editor that isn't geared towards programming. Now, I need to try to align the stuff by literal spaces and make my indentation with them, even though that's absolutely not what spaces are for. Tabs are literally for indenting. They inherently group things together. And tabs are more universal. If you open a file in an editor and hate how much it indented things, just change the tab length it displays. Changing the length of soft tabs in an existing file isn't something as universal. Also, tabs save space on your files. And sure you could say you chose the number of spaces to structure your code a certain way, but imo tabs do the same thing. In fact, I'd say that's what tabs are actually made for, unlike spaces. Edit: typo

[@jiangshenghai57]

Great work, Jeff! Was trying to find the solution to solve this x86 images running on ARM for days. This is exactly what I need

[@grahamc7248]

“vim > emacs” ...“I’m gonna use nano” 🤨

[@greatestsonghits]

I love these videos! Learning soo much. Thanks Jeff!

[@williamkmanire]

The best version of vim is implemented in Emacs lisp.

[@RayStatham]

Excellent video series, really engaging and enjoyable. Thank you.

[@donporter8432]

Wonderful information and timing! Thanks a million!

[@davidhertzberg]

Love your theatrics. Fantastic channel.

[@maartencarbonez1475]

Awesome video as usual! Thenk a lot! But one small comment about accessing the minecraft server: when you deploy a service in loadbalancer mode, you should be able to access it via the service, in this case: port 25565. No need to use the NodePort for that (normally...).

[@julienbietlot3401]

Excellent video Jeff, well done !!!

[@chetana9802]

I wish you could make one video or blog post on all the limitations of such a cluster in an production environment for wordpress or anything else like that

[@AnaPaula-so7ws]

Hey Geff, or should I say Jeff? Thanks for this series! That's really helpful!

[@mil3k]

Do you mean Geoff?

[@hemantghuge7741]

Thanks for sharing lot's of information. Can we do object detection using clustering? Running object detection code on two or more RPi 4 device using k3s

[@Lougehrig10]

0:30 Im sorry, but tab is a dedicated whitespace character just like new-line or space. Just like how new line is supposed to give enough space for a new line, and space is supposed to give enough space to show a separation between characters, tab is supposed to give enough space to show a separation between concepts. Just create a standard of how many spaces a tab is supposed to be and use tab.

[@amrsoll]

This is also exactly what I need :D i get tons of raspi 3s from work because they get their i2c bus wrecked. I needed a kubernetes project to leverage all those low power consumption cpus!

[@MatthieuCoder]

Dig straight down!!! Go for it

[@JeffGeerling]

🔥

[@lawrencelamprecht3180]

Jeff I am very sorry if this is long winded, but I have been fighting with learning about clustering on the RASPi 4 with 8Gb RAM for the last 5 weeks and as you have said is the case ... I feel overwhelmed. I intend to have 7 pis running the cluster all with 8Gb RAM ... so I am sure that it should be a good project. I have managed to get the cluster up and running but until I stumbled across your channel, I only have the following namespaces running. below is what I have installed, but no matter what I try I cannot see if Prometheus is working or how to view the monitoring. This is most definitely because I am not fully aware of what I am looking at and what I need to do to view the monitoring. I am considering dumping the whole lot because for one I am using the 32 BIT OS, I would like to change to the 64bit. kubectl get nodes -o wide NAME STATUS ROLES AGE VERSION INTERNAL-IP EXTERNAL-IP OS-IMAGE KERNEL-VERSION CONTAINER-RUNTIME node1 Ready master 25d v1.19.9+k3s1 192.168.1.241 Raspbian GNU/Linux 10 (buster) 5.10.17-v7l+ containerd://1.4.4-k3s1 node2 Ready worker 25d v1.19.9+k3s1 192.168.1.243 Raspbian GNU/Linux 10 (buster) 5.10.17-v7l+ containerd://1.4.4-k3s1 node3 Ready worker 25d v1.19.9+k3s1 192.168.1.247 Raspbian GNU/Linux 10 (buster) 5.10.17-v7l+ containerd://1.4.4-k3s1 node5 Ready worker 6d22h v1.19.10+k3s1 192.168.1.248 Raspbian GNU/Linux 10 (buster) 5.10.17-v7l+ containerd://1.4.4-k3s1 kubectl get pods -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES prometheus-monitoring-kube-prometheus-prometheus-0 2/2 Running 9 2d5h 10.42.4.25 node5 monitoring-prometheus-node-exporter-drwb6 1/1 Running 7 6d22h 192.168.1.248 node5 monitoring-prometheus-node-exporter-j84j6 1/1 Running 0 8h 192.168.1.190 node4 monitoring-prometheus-node-exporter-fcvt7 1/1 Running 13 8d 192.168.1.247 node3 monitoring-prometheus-node-exporter-srnwl 1/1 Running 20 8d 192.168.1.241 node1 alertmanager-monitoring-kube-prometheus-alertmanager-0 2/2 Running 18 8d 10.42.0.185 node1 monitoring-prometheus-node-exporter-vf6tn 1/1 Running 10 8d 192.168.1.243 node2 kubectl get namespace NAME STATUS AGE default Active 25d kube-system Active 25d kube-public Active 25d kube-node-lease Active 25d kubernetes-dashboard Active 24d minecraft Active 23d I used k3sup to install the master and join the nodes. I was also shown something called k9s, also a great view of the cluster. I would like to get the prometheus working though. I am contemplating deleting all namespaces that have prometheus pods running on them and then following your instruction to get prometheus up and working. Thanks a million for a great channel. I have registered and will stick around. Thanks Lawrence

[@blkfngrs636]

great work! also waiting pi 4 card pricing to see price/performance options for this. probably a really dumn question..i understand adding more pis to a K cluster as you explained. Does the turing pi board support chaining 2 or 4 boards of 7 each. i looked thru the turing spec, and did not see a clear chaining or bus extension method except the i2c port extender? thanks

[@mpwsh]

Found your channel today. Excellent content!. I'm working with a kubernetes cluster in vagrant but it felt wrong to have all the nodes in the same machine, so i'm about to pull the trigger on 4 raspberry 4b to run it. Your findings will help me tremendously, so thank you!

[@JeffGeerling]

If you ever do want to try to get it running on Vagrant, I have a setup in my Pi Dramble repo that I've used for local testing in the past: github.com/geerlingguy/raspberry-pi-dramble/tree/master/testing/vagrant

[@henryolson9732]

I'd be interested to know your thoughts regarding shared storage for the cluster. Is there a best way to share a single high-capacity USB drive amongst all compute units?

[@virtualtools_3021]

share it over the network, theres a lot of ways you could do it, you could use a router with a usb port, you could share it through your OS of choice sharing utility etc

[@TweakMDS]

Great video - subscribed! A bit of constructive feedback if you're looking for it... It would be nice to tie some things together. For example: showing the pi-hole prometheus exporter, so you have a single monitoring source. Another nice thing to demonstrate is avoiding nodeport and using different loadbalancer IP's using metalLB (and again, showing those in grafana). We are probably very well educated with either wordpress or drupal, but it could - in this case - be very educational to show them using a shared mariadb instance, maybe even with failover. I think the minecraft part deserved it's own video. I mean this in the best way, I've been doing a lot with K8s over the last year, and had this been around a year ago, it would have saved me tons of time.

[@JeffGeerling]

That's the pain of video content-I literally have an extra 4-5 hours of content I _could've_ put in this video (and I also did a few things like 4 different ways but you only see the one way I did it in the video), but it's meant to whet your appetite and hopefully get viewers to go the extra distance. Things like MetalLB in particular can confuse the heck out of a K8s beginner, so I don't want to try to introduce it at this basic level. I'm considering doing a 'Kubernetes 101' series later on (like my 'Ansible 101' course), and I'd get deeper into all these topics. So stay tuned, and thanks for the feedback!

[@GreySectoid]

Tabs in the beginning of line, spaces in statements. That way you can set the tab width depending on how much screen space you have available but it will still always align special cases (e.g. multi-line statements, tables, comments, ascii graphics...) correctly. On small screen I've used tab width 2, but on larger screens I can use 4 or 8. There's no reason to fix that at write-time by using spaces for a purpose tabs were created to begin with. The flame wars around this issue are only because people haven't yet gotten into the habit to write tab-width-independent-but-still-correctly-aligned code and insist using either-or when the correct answer is both.

[@nyny]

A video showing how you keep something like this stable over months, or long term would be nice. I set these things up, and when I go to do something a month later, everything is out of date and updating becomes such a mess that I end up tearing everything down and restarting from some new tutorial. That is part of the fun of learning, but at the same time, I'd like to have some reliability like what I get with say my synology.

[@johnm2012]

Jeff, I noticed you switched to the root user on the master node early on and stayed there for the remainder of the video. Was that just a simplification to avoid adding the complications of permissions to an already complicated subject, or does it all really have to run as root? Just wondering what is best practice.

[@JeffGeerling]

It was a simplification; I think I mentioned that there are other ways to do it, and for a couple of the K8s configs too, there are more secure ways to configure things-but for the purpose of instruction, it is easier to mention the fact and leave out the detail (lest this video go on for another few hours 🤪). I recommend *not* performing every action as root, for various reasons, but instead to make use an unprivileged user, set up `kubectl` by copying down a KUBECONFIG to a non-privileged account, and even after that, using a service account/user that only has the privileges required to do what the user needs to do (e.g. manage one namespace, or just deployments, etc.). But those are some things that I may be covering in a future series on Kubernetes; in the case of the Turing Pi cluster, it's much easier and is a more consistent configuration to log into the master Pi and switch to the root user (since all the `kubectl` commands will 'just work' that way).

[@johnm2012]

@@JeffGeerling Thanks for the detailed answer and thanks for making this complex subject more accessible.

[@tdiallo199]

amazing tutorial. Thank you Jeff!

[@paintenzero]

Hi, Jeff! Your video is super cool! I enjoy it a lot. I have one question: what do you use for storage backend? Did you find something that works well with ARM cluster? K3s default is to use local-path PV that kinda binds pods to the node and prevents it from running on another node even if the first goes down. Also it doesn’t replicate data in any way. I would like to hear how you do it. Actually storage is such a big topic in Kubernetes that I think it deserves a separate video. As for me I found it easier to not use storage backends at all and just create glusterfs across all the nodes and use local directory that points to glusterfs mount. This way pod is not bound and glusterfs replicates data between nodes if you configure it right. But that is definitely not a Kubernetes-way to do stuff but I just could find storage driver that runs on ARM :(

[@mnieri71]

It worked so smoothly. Thank you! I would be interested on how to use the cluster to scale the applications. For example: how can I scale the db (MariaDb) behind Drupal

[@Dome123456]

Not all applications scale like that. MySQL is a good example. Basically you can’t scale it without forming a database cluster (which has nothing to do with kubernetes) or replication. In theory you could do that inside kubernetes but that would involve a very complex configuration management, orchestration and health monitoring behind it and in most cases, simply not worth it. Rather build your database cluster outside of containers and learn before even thinking about doing that inside kubernetes.

[@Fractal_32]

1:30 makes me think of me running Arch Linux, at first you may not understand how to do something but when you fail you learn very fast how to do it correctly or how to perform the task a better manner.

[@LuisNovoaRomo]

Hi, Jeff, a great series of videos, thank you for all the effort. After many tries to install Kubernetes (microk8s and kubeadm) in my R-Pi cluster (2 nodes 3B and 2 3B+) finally I started to work with Swarm, I didn't try k3s because one of the services that I like to run in the cluster is Pi-hole and I didn't find information about to run Pi-Hole over container-d, but now with this video, probably I'll give an opportunity to k3s. Currently, I'm working to create (and learn how this network works) a macvlan network for the Pi-hole, because I want to Pi-Hole offers the DHCP service to my network. I found how to do it in Swarm (after many essences similar to yours about frustration), and finally, I have an idea of how it works. My question is if in k3s is possible to have the same configuration and have a "virtual NIC" with its own MAC address to have "direct" access to the network and have the service DHCP of Pi-Hole available in the network. Nowadays I'm in the process to update my knowledge about technology, learning Docker, Swarm/Kubernates and other stuff. Thanks again for your videos and for your advice.

[@squeakytoyrecords1702]

@Jeffgeerling Hi, thank you for this wonderful video. My question is, can I use Math Lab on this cluster?

[@jnoe491]

Videos are great i have my own cluster built right now and i have honestly ran into many hiccups and over a few days thinking about it, it was all solved. Anyone interested in doing this should def try it out.

[@RichardBronosky]

45:40 "I'm going to grab my values" really doesn't help up. Where did you even get the template for that file?

[@misaalanshori]

Okay so my knowledge with kubernetes and clusters is little to none, but i didn't know you can run a Minecraft Server on a cluster... So how does that work? So is the minecraft server running on multiple compute modules and then showing up as a single server to the player? I know you can put the minecraft server in a container like docker, but then how does it work in a cluster? Or is there just one minecraft server running in one of the compute module of the cluster? And kubernetes is just a way to deploy it to one of the compute module in the cluster? Usually when i think of someone deploying something to a cluster i just immediately think it will be running in parallel and basically combining the power of multiple computers to run a single program.

[@thomaspreissler6476]

Jeff, many thanks for this video, it's great. Two questions: I got a "CrashLoopBackOff" for "tiller-deploy-6974685dbc-56vnr" after "helm init". The logs say ... "exec format error". How did you get around that? And then you easily tie Grafana for example to worker-02. I just restarted mine ... and K8s has changed the node of course. How did you 'tie' it to worker-02 in your setup, it appeared to be working for you?

[@thomaspreissler6476]

First Question: I found a tiller-deploy for armhf further down. To get it deployed, you need to do a kubectl delete all -l app=helm -n kube-system to remove everything regarding helm, in my case this also removed the CrashLoopBackOff pod due to exec format error. Then deploy github.com/jessestuart/tiller-multiarch with kubectl apply -f manifests/tiller-rbac.yaml helm init --tiller-image=jessestuart/tiller --service-account tiller and tiller-deploy is running ok after that. Second Question: Short answer - it doesn't matter. In this demo instance K8s will route the request internally to the worker it is runing on. Look it MetalLB / Traefik if you want to feel some pain.

[@martinsmee23]

Hi, very much enjoying this how to. Now that the cluster is working, can you do a video on the correct steps to remove one of the Pi's and show how the cluster handles this. Then steps to show how to add it back which will also demonstrate how to expand an existing cluster. Many thanks.

[@DoorThief]

I love that Silicon Valley clip! Tabs > spaces.

[@davidsonmg]

Chuck sent me! Sooooo much to learn!!

[@bdollerup]

Another great tutorial. I just put Prometheus and Grafana on my k3s cluster and got a lot of great information out of that. Doing so did raise a couple of questions, though. How would I add additional servers to be monitored by the Prometheus instance in the cluster and what about servers running in Azure and AWS: How would I add those and what are the security implications of exposing an HTTP endpoint publicly?

[@maxbernard1393]

Hi Jeff, thanks for the great tutorials. I'm having a bit of trouble with this one. At the point where you get the ingress URLs, if I paste any of the URLs into my browser, I get a 'connection refused' error message. Any idea what may be causing that, and how I might go about fixing it?

[@uruemuesiridjebah8628]

I'm getting same error too. how did you eventually get it to run?

[@maxbernard1393]

@@uruemuesiridjebah8628 Unfortunately, I never did. I eventually decided to wipe everything clean, and start over, using an X86-based cluster, instead of the Raspberry Pi, and that worked much better.

[@uruemuesiridjebah8628]

Alright then, thanks. @Max Bernard

[@ryanjenks8771]

Teacher, may I be excused? My brain is full after watching this.

[@DeepakRamanath]

Quick question. I'm new to Kubernetes and I do not see that you install docker-ce. So wondering how is k3s managing?

[@StewieWilfred]

Awesome video! Not sure if this is discussed, but is there any reason i wouldnt be able to use different generations of raspberry pi in the same cluster? Want to avoid having to buy more!

[@smartassist9700]

Large ice milk b4 bed ... it’s 245 am.lol. I love your videos.

[@asimami]

Hi Jeff, Amazing content as always. I'm waiting for my turing-pi to arrive so preparing for the eventual cluster adventure and your videos are very helpful. My questions are more related to the apps you run on your mac. I wanted to know what was the app that told you the IP address and had a network graph running in the top bar. I would also appreciate if you share other developer-productivity-enhancing GUI/taskbar apps for mac that you've found useful.

[@makonnen2112]

Greate video! Just wondering if you could increase the cluster memory by mounting an NFS file system to a NAS, what do you think? Cheers mate!

[@JerryOmann]

Which storage class do your Turing cluster use, so that your Minecraft Map is available if one Pi got killed?

[@tommytigerpants]

Hey Jeff, it would be great if you could do one of these again as I cannot find any guides that use the latest versions and resources. It's really challenging to do this much configuration as an amateur without the guide. I have attempted to build this setup (following a mixture of this guide, Network Chuck's and other smaller KZbinr's guides) only hit wall after wall. I know you're busy and this may be a low priority, however it feels like this style of RBpi usage is losing favour or becoming less popular. I wonder if it's to do with the cost of the modules? Anyway, I'd really appreciate it as I have been stuck for a about 18 months now! Cheers

[@GanonMasta]

Thanks for the Ansible playbooks, I'm sure they will come in handy. But damn, I was looking forward to see what the performance of a cluster like this would be!

[@JeffGeerling]

Don't worry, that's coming next episode!

[@GanonMasta]

@@JeffGeerling Looking forward to it! I'm on the fence about building myself a small cluster of pi 4's. I'm definitely excited to see an overview of what kind of performance to expect!

[@sgreszcz]

Jeff Geerling great episode. Would be interested in knowing the ‘best’ os for running k8s/k3s as far as available container images, etc. I need to compare ubuntu64 for arm, hypriot, etc.

[@nicktipton5675]

Hey in the original disney animated Aladdin movie, does anybody remember what the name of the charecter that Robin Williams played was? My neighbors pi-hole is blocking my imdb access....

[@larsskage5584]

As always, I learn a lot from your material. One question - are there any benefits or drawbacks to cloning the GH repo to the master and run kubectl there compared to on a remote machine? Looking forward to the next video so I can continue evolving my K3S cluster of four RPi4.

[@der_quarks_pro2462]

A few questions I have: 1. Does the minecraft server run on one single 'worker' or is it processed by multiple 'workers'? 2. Is it ever possible to proceed one task (e. g. a running minecraft server) by multiple 'workers'?

[@JeffGeerling]

Minecraft server is a single-threaded application, so unfortunately you can't run it in more than one Pi and 'scale horizontally'. For other applications, though (like Drupal and Wordpress), you can scale the apps by setting `replicas: 2` in the Deployment, and the Kubernetes Service will balance requests to the apps that way. One note, though: you have to change some settings and use some sort of persistent storage layer that allows the PVs/PVCs to be mounted by multiple containers at the same time. I usually use NFS for this purpose, but Ceph/Rook are also popular.

[@exoprinter1373]

3:29 - when you look like an NPC strait out of Fable trying to run away.

[@venaist]

Your videos are gold!

[@roguethinker6284]

This is a fantastic channel!

[@GoncaloSantosHit]

We can use qemu for building for ARM on x86, my desktop has 32 threads, i use it to build images for my Pi´s.

[@michaelwallis1130]

Hey Jeff ... I'm running on an old iMac and with just 2 RPI systems, but when I get to the "kubectl apply -f manifests/setup" stage, I get: "-bash: kubectl: command not found" ?? I can't find an executable "kubectl" at all. :-(

[@EndermanAPM]

My problem with k8s is how to handle persistent storage with multiple instances. For example, what happens if the node with the persistent storage dies?

[@OpletalRobin]

Any idea why my nip.io wouldn"t resolve?

[@hfh_hfh]

To do with the way your k3s cluster ingress bridges outside the HTTP/HTTPS traffic to the pods

[@mattk3035]

Thanks for the video, however somehow I'm stuck trying to deploy the chart for pihole. Any help would be appreciated, thanks in advance! I get this error "Error: Kubernetes cluster unreachable: Get "localhost:8080/version?timeout=32s": dial tcp 127.0.0.1:8080: connect: connection refused" I can run kubectl commands fine, nodes are up and ready. From what I read helm should be using that cfg that kubectl uses. I have tried several versions of the install string but no luck. He's the one I think is right. 'helm install --version '1.7.20' --namespace pihole --values pihole.yaml pihole mojo2600/pihole' running from ssh on my master from windows powershell.

[@user-be3mk6cg4c]

I know this is 9 months later but i had same problem and solved it with by running export KUBECONFIG=/etc/rancher/k3s/k3s.yaml Found it on the blog, thanks Tugay!

[@Drag0nvil]

I run my Minecraft servers in a container... However it is modded Minecraft and I think it should not be a use case for the cluster. My headless server had an i5 3337u and 16 GB memory. The bottleneck I experience is that it uses up 100% of one of my threads. I don't think Minecraft would profit from a cluster with a lot of cores since it seems to be single threaded. Also can cubernetes split a container over multiple machines? If I'm wrong on this please tell me how to unlock multi threading and increase performance :)

[@paumb64]

minecraft is single threated, that's it. the best way to improve performance is to use fabric and install these mods on the server: carpet, lithium and phosphor. all of them increase performance, much more than paper

[@inachu]

Familiar with the Pi but not the cluster. Can Windows IOT run on a cluster?

[@DaPanda19]

I'm a but stuck on the web UI, I've gotten the cluster monitoring installed with applied manifests, but I cannot access Prometheus or Grafana after running "kubectl get ingress -n monitoring", I see grafana..nip.io but I get get site can't be reached. Unsure of what I messed up :(

[@alexbissessur5013]

Were you able to fix it? I've been trying to fix that for ages now and I'm on the brink of a mental breakdown. I think I had to install nginx, which I did. I tried configuring it but now am just getting a 502 bad gateway.

[@Schlumpfpirat]

So what would happen to the data if the marinadb cluster-worker died - is it distributed equally at all times or how's storage handled? Really good series btw, even though I can imagine setting up SSL could be a nightmare...

[@sethalump]

I thought I recognized your name. Thanks for all the Ansible roles ! =)