Hurry up and wait, or why as a sysadmin I don't do shit for new deployments until there's network...
We're working on deploying a new system that is months behind schedule. The reason? The LLM hype machine and OEM's allocating all their product for hyperscalers and/or companies that want to toss their ring into the shit show. It's unbelievable..
Anyways, for the last month or two, while there was NO NETWORK ACCESS AVAILABLE, i.e. THERE WAS NO WAN/LAN CONNECTIVITY FOR OUR EQUIPMENT, people I work with have been dropping suggestions of maybe I should provision a node. My consistent answer is, "with what network?". Seriously, I get wanting to POST new machines and get an O.S. installed, but we're talking about hundreds to thousands of machines, depending on the cluster. This new system will easily be ~2,000 nodes, not including any VM's that we decide to roll out for sorry ass applications that can't use an entire machine! Back to the point, the look of shock and/or confusion, and/or disappointment, etc., I see on other people's faces when I tell them that I'll start as soon as there is network available. Sure, I could get an ISO, install the O.S., and then constantly go back and forth between a connected machine to download packages, repositories, configuration files, etc. Why though? Why would I want to waste my time going against the grain of typical network provisioning, to have a node "ready" without its fucking network? It makes absolutely no sense at all! Sure, there are those that would assume maybe just use a switched fabric for "connectivity"; up until yesterday, there was no network available at all, not even an "air gapped" LAN.
I know how heirarchy works, and I know that those in the upper crust want "deliverables". However, given months of delays, delivery issues, and no network, there is _NOTHING_ to deliver. I've always been the type that prefers to start with everything - think like assembling your own furniture - I will pull out each piece, verify it is listed in the instructions, and then sort, and then proceed to build. I treat computing systems the same; if the basic building blocks/components are available, then yes, we can always add additional equipment/resources into a working cluster.
I imagine I received that look because there are many steps involved in setting up a system that can provision other nodes. I guess they've (the upper crust) have probably experienced their share of "admins" that are slow and probably overestimated their abilities, leading to self inflicted delays. However, with the power of ADHD, I can use my superpower of hyperfocusing to get days of work done in a few hours. Case and point, network was established on 2026-05-05 at 10:55, and a little over two days later, we are ready!
- DHCP server : check
- TFTP server : check
- HTTP server : check
- NFS server : check
- Support for x86 and ARM architectures : _Fucking CHECK_
- Configuration management : You _fucking better believe it.. CHECK!_
All that to say, no network, no reason for me to do shit with new stuff.
Here's some eye candy
My view from earlier today
mrfusion in a stunning reveal (I smudged the other admin's face, didn't want to ask permission to post)
- LMFAO, someone said it was "Shark week"
More bitching and complaining
I want to go home