toad.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
Mastodon server operated by David Troy, a tech pioneer and investigative journalist addressing threats to democracy. Thoughtful participation and discussion welcome.

Administered by:

Server stats:

297
active users

#ipxe

0 posts0 participants0 posts today

Let's say you run a public TFPT/iPXE server that will dynamically compile images to be worker nodes.
You need to uniquely identify each worker node in the built image, let's say hostname has to be unique and baked in.

What would you use?

You can't rely on IP cause they might be behind NAT, and you can't use MAC cause ARP and multicast for ND/NS aren't publicly routable.

Ideally if the same client returns for the boot image you don't just give them a new hostname/identity.

@SweetAIBelle very nifty...

Ideally with some nifty tools (compareable to Tails' #PersistentSotrage) to allow mounting local/remote storage as $HOME and potentially leverage toram as boot option.

Certainly that "#RescueLinux" part is mid- to ongterm goal of OS/1337 and I am working on it (abeit at a pace slower than I want to cuz 'things happen in meatspace' and those sadly require my full attention...

  • Needless to say it's always good to have a nifty system at hand (worst-case I'd install @ubuntu / #UbuntuLTS #Desktop on an external SSD/HDD and have that at hand) if one doesn't have their Distro on a hot-swappable / toolless replaceable drive anyway...

Using #ArchLinux is quite good because it makes it easy to get started and adapt said system as need be.

Replied in thread

@dalias @eniko Personally, I'll sooner or later have to engange the whole #ARMv5 / #ARM11r7 / #RaspberryPi architecture anyway with OS/1337.

  • Tho given it's a minimalist Linux merely booting an initramfs and never doing any writes, there should be no indurance issues.

But I know a #readonly - OS isn't practical and where it is people already use #iPXE & #iSCSI for #diskless setups!

  • THO AFAICT there should be an option to basically partition the system and only keep /boot on an SD card, and the rest could be on any USB mass storage device (i.e. SSD or even HDD)...

That being said Raspberry Pi do have the key advantage of being by far the best in terms of #documentation.

Replied in thread

@http @Rob298 precisely that!

  • Also most corp/org/edu networks only backup the $HOME directory and sometimes even allow syncing them across distros & keep them across version updates, so all the settings, addons and stuff remaib where they are: in said /home/ subfolders!

  • In fact most places with a sizeable #Linux-#Desktop landscape will just keep the /home/ directory on a redundant, #iSCSI-SAN and #netboot their #DisklessWorkstation|s via #iPXE, as this way burglars stealing devices most likely end up with a locked-down machine (anything but booting the preset network targets won't work without admin password!) that is a paperweight to them and espechally no data, which is crucial when it comes to #ITsec, #InfoSec, #OpSec & #ComSec.

Cuz it's way easier to secure 1-5 server rooms than thousands of publicly accessible machines on multiple campuses.

  • It's also a real godsent for technicians as they can just login / boot into a diagnostics system and quickly see what's wrong if a system has issues (if they don't already see it in their dashboard that collects logs, telling them they need to replace the CPU fan or clean a machine as it's overheating)...
Replied in thread

@uastronomer it's something I.did implement in the past (abeit #KVM + #Proxmox, but the steps are similar enough):

You can seperate #Storage and #Compute given you have a Storage-LAN that is fast enough (and does at least 9k if not 64k Jumbo Frames) and have the "Compute Nodes" entirely #diskless (booting via #iPXE from the #SAN) and then mount the storage via #iSCSI or #Ceph.

  • Basically it allows you to scale Compute and Storage independently from each other as they are transparent layers and not be confined to limits of a single chassis & it's I/O options...

Did a bigger project (easily 8-digits in hardware, as per MSRP) where a Employer/Client did #CloudExit amidst escalating costs and #ROI being within quarters (if not months at the predicted growth rate)...

Replied in thread

@Saupreiss Klar sollte für all solche Fälle entsprechende Contingency existieren...

Sprich ne Möglichkeit jene Systeme per #iPXE automatisch auf nen Rescue-System zu booten um dann nen Image draufzuschubsen...

  • Wenn nicht dann haben diejenigen die's Betreuen nicht ihre Arbeit gemacht!
Replied in thread

@cdp1337 personally, I've wasted 15 years of my life trying to make #Windows useable and the main regret I have is not switching to #Linux 100% 5 years earlier...

  • I can't do Windows at all because no amount of "pain money" will compensate for the damage it'll take to my mental health.

I'd rather not do #IT at all than suffer through Windows.

  • And yes, shit's that slow and hard, I've seen companies buying ~ €200k (MSRP which obviously noone pays) in low-core count EPYC servers just to have #MSSQL be somewhat reasonably fast.

Not to mention #Microsoft's #Licensing is just absurd as they demand that businesses license their garbage not for the Cores/Threads they'll assign to the Software, but for every single core the #Proxmox or #ESXi cluster has, even if it's impossible for said Software to run on those at the same time because "#WindowsServer" doesn't even support multi-node clustering or fault-tolerance at all.

  • I could at least reason this as a price to pay for it if it actually bought something to the table worth it...

But it just doesn't!

And yes, Windows does require reboot orgies to update...

  • Why else do you think it's less pain and frustration to use #nLite and #iPXE to build unattended install images and just reinstall all Windows #Desktops everytime a #ServicePack gets released?

It's way faster that way.

  • Also Microsoft's own tool for that, #WAIK, really sucks ass and is a bigger resource hog than a "Full Installation" of #VisualStudio Enterprise Edition!

Again: I'd rather spin my own distro from scratch and daily drive it than ever touching Windows again!!!

Veracious NetworkCharlie (@cdp1337@social.veraciousnetwork.com)So serious questions, (as a tangent from today's events)... I use #Linux daily and have used it for decades now; for desktop/gaming, specialty appliances, and of course servers. Updates are simple, quick, and usually painless. Over this weekend someone asked me to help with a #Windows box. New install of Server Standard; can't activate. I have a key, but the damned box refuses to talk to Microsoft activation servers to let me enter the key. Also updates... ONE single update took about 4 fucking hours to install! It downloaded quickly, but the update screen just SLOWLY ticked up percentage by percentage. Is this normal?!? 4 HOURS!!! This is on XEN/Proxmox, but I've experienced HORRIBLY slow update installs on bare metal too. And then it comes to #Exchange; I've been trying to install an update for that but every damned time it just spits back 'You have a pending restart...' According to the help docs, "If you see this error after you've restarted the server, try restarting again" WTF, I thought that was just us Linux folks razzing on Windows. Does that OS seriously require several restarts just to install one fucking package?!? Is there some GUI option labeled with "snail speed only" checked that I'm just missing?!? #HomeLab #Admin