Member
We run our own scrubbing in Frankfurt and Amsterdam. Cloud scrubbing is fine until you get a persistent attacker who learns your provider's nullroute timing. Anycast with regional sinks, 30 second shift on attack signature change. Six figures doesn't buy you out of 3am calls though.
30 seconds? I'm jealous. My provider needs a ticket for anything manual. I moved three status pages to UptimeRobot this year because the native ones lie. Green on the page, red in reality. Customers notice before I do.
Member
ARM cores don't wake you at 3am because they're efficient, not because they're ARM. My Rockchip cluster has been up 8 months without intervention. X86 thermal throttling at 2am because the datacenter HVAC cycles, that's your real enemy.
Member
This thread is overwhelming. I just got my first VPS last month, Hetzner, and I broke the ssh config in 20 minutes. Had to use the rescue console. Is lying really the way in? I can learn bash properly if that's what it takes.
Member
No. Lie about having done it, then do it immediately so it's not a lie anymore. That's the trick. I said I knew Proxmox, spent two weekends breaking it, now I do know it. The lie is just a deadline you set yourself.
Member
My first VPS was 97.3% and I thought that was good because it was cheaper than my coffee. Now I run two in different cities for $7 total, failover with keepalived. Still cheaper than coffee.
So I should break my CloudCone on purpose? I already do that by accident. Maybe I need to document it, put "incident response" on my resume. This is actually helpful, thank you.
CloudCone at $3 is overpaying. I got a 2GB yearly from RackNerd for $18 last Black Friday. $1.50/month if you do the math. LA datacenter, same as CloudCone. You're burning money for the brand name.
Member OP
I put "performance optimization" for fixing sv_minrate because the server choked on 32 players. It was one cvar. Interviewers don't ask details, they ask what you learned. Say "I learned that defaults assume the worst hardware." True and impressive-sounding.
Member
I rewrote our log parser in Rust, 40% faster than the Python it replaced. Still runs on x86 because that's what the datacenter has. ARM migration is a procurement problem, not a technical one. Rust runs everywhere anyway.
