HomeLab *anywhere* using Kasm Workspaces simply with web browser

[hwzlite]

Look mum , clientless access with just Web Browser

Replacing my Apache Guacamole in my adventuring into proxmox.
A well polished s/w offers much more other then the standard vnc,ssh,rdp support, which can spin any available apps in Workspace Registry

• live demo on sprining Ubuntu Jammy : https://kasmweb.com/#dialog-2

 

[hwzlite]

Aside install the full-fledged Kasm Workspaces , you can also easily spin its standalone container image in your existing docker infra and access (via https://<your ip>:6901) through a web browser too.

Do note that this lightweight alternative comes with this limitation: "Please note that some functionality, such as audio, uploads, downloads, and microphone pass-through, is only available when using Kasm Workspaces for orchestration."

Example: Checkout Launch a real-time demo for Chrome

• Kasm Depositories: https://hub.docker.com/u/kasmweb

 

[CoolRock]

bookmarked to read later

 

[Deathzz]

what infra are you running proxmox for kasm and misc? does a mini pc with n100/200 work well enough? or we need more modern chipset?

 

[Phen8210]

Interesting, but the main reason I run Linux to host my api servers is because those VM I rent have potato spec, no choice.

If the machine is powderful I might as well use windows, but also depends on how powderful it is.

 

[hwzlite]

what infra are you running proxmox for kasm and misc? does a mini pc with n100/200 work well enough? or we need more modern chipset?

Ok here the stats, currently running these VMs full-time:
OpenWRT - 4 vCPU, 512MB RAM (softRouter for M1 500Mbps broadband with Cake SQM)
Win11 - 4 vCPU, 6GB RAM (HTPC on physical GPU passthrough to HDMI > Monitor)
CasaOS - 4 vCPU, 7GB RAM (Docker containerization running Kasm, AdguardHome, ntopng, Tor, qbittorrent, cloudflared...etc)

As you see, very stretched on RAM usage liao, probably will upgrade (from16GB) to 32GB DDR5 RAM in the near future...

Ideally, Kasm should run on a dedicated VM itself, but I run it as a docker container in CasaOS instead.
For some weird reason i encountered error when try to deploy Kasm image from CasaOS-LinuxServer-AppStore , but goes well on big-bear-casaos .

 

[Mach3.2]

very stretched on RAM usage

Yeah for virtualisation, you're mostly limited by RAM, since overcommitting RAM will lead to disk thrashing, which sucks for performance.

You can overcommit CPU by quite a bit. pCPU to vCPU ratio of 1:2 or 1:3 is common. Mine is just a little over 1:2.

 

[TanKianW]

Yeah for virtualisation, you're mostly limited by RAM, since overcommitting RAM will lead to disk thrashing, which sucks for performance.

You can overcommit CPU by quite a bit. pCPU to vCPU ratio of 1:2 or 1:3 is common. Mine is just a little over 1:2.

Dun we all just love to overcommit......our CPU (just make sure not your relationship)..

 

[Elijahonli]

LOL i overcommitted mine to 1:3.66

 

[Mach3.2]

overcommit

that's why we virtualise

i overcommitted mine to 1:3.66

I think it's still ok if the other VMs idle most of the time. if most VMs are idling, you can probably over-commit CPU by quite a bit. I think the way people do it is to keep CPU ready time below a certain threashold, no so much on the actual over-commit ratio.

 

[TanKianW]

that's why we virtualise


I think it's still ok if the other VMs idle most of the time. if most VMs are idling, you can probably over-commit CPU by quite a bit. I think the way people do it is to keep CPU ready time below a certain threashold, no so much on the actual over-commit ratio.

I will normally run 2x/3x identical production servers with shared data storage (or snapshot) on a separate network storage node/location on TrueNAS/Ceph. Let HAProxy on pfsense be the load balancer (or a dedicated LB-er) to handle the traffic/load (or switch) between servers while XOA handles the "high availability" when either VM are down.

The beauty of "virtualisation"! ​

 

[hwzlite]

that's why we virtualise


I think it's still ok if the other VMs idle most of the time. if most VMs are idling, you can probably over-commit CPU by quite a bit. I think the way people do it is to keep CPU ready time below a certain threashold, no so much on the actual over-commit ratio.

Worse comes to worst for Proxmox can further tame with "cpuunits (nowadays often called CPU shares or CPU weight)" to prioritize "important" vm

 

[xiaofan]

You guys really run heavy stuff. I do not need to access the home network stuff outside, other than to access the web GUI of the two routers (two networks: Asus RT-AX86U and OpenWRT on Intel J4105) to see if they are having issues. I use Asus Instand Guard and Wireguard VPN server on the two routers to access the router.

My "home lab" (if the setup can be called Home Lab) consists of two always-on Intel mini PCs. No NAS. No Windows VMs. No home automation stuff. You can see no problem with the CPU/RAM usage at all.

I have a few other Mini PCs and two Raspberry Pi 400 but they are usually OFF and only ON when needed, just like my Acer Windows 11 laptop and Mac Mini M1.

1) Intel J4105 mini PC (8GB/256GB) with quad Intel i210 NICs, running Proxmox PVE 7.4. It runs my home network with OpenWRT (or pfSense or OPNSense) and Adguard Home / Pi-hole containers. It has a few other Linux/BSD VMs as well but I am not really using them much now. 8GB RAM is the main limitations. I will run only one VM at a time here.

This may be replaced by a new Intel N100 mini PC (16GB/512GB) with quad Intel I226-V NICs.

2) Intel N100 mini PC (16GB/512GB) with dual Intel I226-V NICs, running Proxmox PVE 8.0. I have about 20 different Linux/BSD VMs (not counting OpenWRT/pfSense/OPNsense/IPfire) installed where I test different Open Source projects. I also play with OpenWRT/pfSense/OPNSense here (Double NAT after the Asus RT-AX86U). The VMs are usually OFF so you can see the CPU and RAM utilization is low. I can easily run two VMs at the same time.

 

[hwzlite]

Another nice use case demo using Kasm Workspaces in open-source intelligence (OSINT) :