2200 LXC Containers on a Mini PC!

[@staticalmo]

Somebody stops this man before he creates a containers singularity

[@marcialabrahantes3369]

stop* 😉😊

[@MarkConstable]

I'm surprised to learn about the 1024 limitation of the native Linux bridge. We wouldn't know unless you did something crazy like this experiment, so many thanks for that.

[@romayojr]

2,200 LXC containers on a Mini PC is insane!!

[@964tractorboy]

The results from this LXC and the previous VM experiments are just jaw-dropping. They do make me wonder what sort of work-loads they're capable of sustaining...

[@druxpack8531]

I really enjoy your videos because you are a systems guy that produces high quality content about true homelabbing (from an IT Pro's perspective)....in that i mean a homelab that is designed around upskilling and not about setting up the rr's to get things for free. Keep up the good work!

[@Polarzbek]

This is fascinating. My takeaway here is no, I don't need another piece of hardware, I'm good to throw another VM, or LXC, or two or three on my existing hardware! Cool stuff, thanks Brandon!

[@visheshgupta9100]

This is absolutely crazyyyy!! I would love to see you do a video on how to run Wordpress in LXC container using Turnkey Linux & a full LAMP stack VM with Apache configuration. The way you explain things are precise and easy to follow. Kudos and keep up the great work.

[@VirtualizationHowto]

Thank you!

[@JamesGreen-gv4yn]

My gut feeling is that you hit the max number of ports on the interface. Leading me to think you could have added a second or third interface and split up the node across them. Avoiding the nested PVE method you used. I am guessing that you could even add "virtual" interfaces on a single NIC (assuming the system has only one) and create a bridge on it. It would be interesting to see if you can get that to work.

[@kingneutron1]

I would also be interested to see if this 1024 limit applies to e.g. a 2-node cluster with Qdevice

[@csgrullon]

I wonder if the limit can be worked around by creating an additional bridge if the mini pc have an extra nic.

[@ronm6585]

Thanks.

[@MarcusWillberg]

I wouldn’t call running hundreds of VMs or thousands of LXCs a ”realistic workload” when they are all from the exact same base image. You do know that the reuseable memory you get when running even slightly different OS and software is not even comparable? Maybe do another test with all different operating system versions and different applications for comparison?

[@jcarman]

I wonder if this has something to do with TCP connections; maybe there are 2048 connections?

[@homemedia4325]

Virtualisation inception... how many ProxMox can run in ProxMox can run in Proxmox ... ... ... etc... and will they cluster... hahah :)

[@geneann8286]

You could try run more virtual PVEs and more LXCs if you use Alpine Linux LXCs with , let say, 128MB of RAM instead of Ubuntu with 512MB. 2200x4=8800 LXCs, possible? I know it is crazy.... Good video though !!!

[@RichardBuckerCodes]

interesting... maybe.... awesome... no. The containers are essentially idle so all that's happening is the host OS is just looping through a list waiting for something to do. And the same can be said for the VM test. Cute, but that's it.

[@Vova67ify]

Has anyone ever seen on this channel a Mini PC 64Gb RAM, 1000Gb SSD and Intel NIC with 802.1Q support? I can't find anything with Vlans support.

[@joshhardin666]

$10k/yr for only 100 machines!? I've got a few hundred virtual machines across a 5 node proxmox cluster... that gets expensive really quick - given how ridiculously expensive $10k is, particularly when there are free open source options for network documentation as well.

[@TheRealAnthony_real]

And I have an lxc container running one docker instance and 50% IO 😅 on a Lenovo v50s with an i7 12 threads .. 80 % CPU .. Obviously his 2k containers are idling only ..

[@XtianApi]

Wonder what the point of all this is? Running all these containers which are collectively not capable of using even the slightest amount of cpu. When you say workloads, what were the workloads? Seems like 10,000 empty word documents, unless I'm missing something. Thank you

[@ewenchan1239]

2200 IDLE LXC containers is, in practice and in essence, quite literally, of no use as none of said LXC containers are doing any (useful) work. LXC containers that aren't doing any work are precisely that -- useless. I think that a better test would've been that if you had set up the LXC container to start LINPACK or something, and then see how many of those that you can be spinning up on this Mini PC -- that would be a MUCH better indicator of what the Mini PC can or can't do. Having 2200 idle LXC containers means nothing.

[@vladchikov6953]

It's some kind of techp0rn)) Thanks for the enjoyment!

[@Hypn0s2]

I would for sure contact Proxmox about this. I wonder if you hit pid_max or file-max in the kernel limit.