• 0 Posts
  • 53 Comments
Joined 1 year ago
cake
Cake day: June 20th, 2023

help-circle



  • Worst is when someone fucked up the DB time configs at some point and you have datetimes in a column that fall during the “nonexistent” hour in which clocks skip ahead for DST, and you have to figure out what the fuck actually happened there, and where in the data pipeline tz data was either added or stripped (sometimes it’s both, and sometimes it’s not just once), and undo the timestamp fuckery.

    Source: did that this week. Was not super awesome.









  • I would recommend:

    • go on eBay and find some sort of cheap Lenovo/dell/hp thin client for your secondary node. You can find workable 1L-class boxes for around $100. You can get away with some of the older m700/710/900/910 tiny models, but the extensibility of the m720/920 tiny models is going to be much better.
    • for your primary, I think you’d probably be best off finding an old server tower with 8 3.5” bays - if you’re lucky and on-the-ball, you may be able to snipe something like this, but shipping is of course going to be a bitch. An alternative is to pick up another one of those thin clients (making sure it’s a model with USB3, but preferably 3.1 or 3.2 whatever the gen is (side note: fuck anyone involved with the USB versioning scheme, because it’s absolutely indecipherable) that can actually support meaningful data transfer, and then just find a cheap DAS and connect it to that node.



  • I suppose it’s a personal choice of where you set the bar for your systems.

    Personally, as a software engineer who’s designed and built a lot of systems over the course of my career: nope - not if you want to just set it and forget it, that is. Which I do. And yes, most of the systems I’ve built professionally aren’t to that standard (mostly due to time constraints), which has consistently frustrated me, but you gotta be a little bit zen about stuff you don’t have complete ownership over.

    Maybe your tolerance for manual intervention is higher than mine, but in terms personal standards, I don’t consider a system to be “done” until it’s configured to to handle itself resiliently and recoverably in all but the most catastrophic situations (I.e. basically, a hardware fault, or some sort of fairly serious upstream infrastructure failure).

    All that said: YMMV. It’s a personal preference, and I know my standards would be considered abnormally strict by some.





  • Lmao dude thats simply not happening at that price.

    You could get part of the way there with an old Dell server, but you’re probably gonna be paying closer to 2-300 for a decently spec’d one like you’re describing. You’re probably gonna be looking at a 10 year old twin quad core setup with a tdp of like 500W combined for JUST the cores. Your power bill is going to murder your budget, even if you somehow find a magical deal on the box.