So you might think that I'd turn this into a database joke, but it gets at something more interesting: the philosophy of failing over. A thread informed by Ancient Sysadmin Wisdom!
https://twitter.com/_msw_/status/1420178242065952769
Back in the Olden Days of Datacenters, when sysadmins knew C, "serverless" meant that Dell hadn't shipped an order on time, and "Kubernetes" held as little meaning as it still does today, the question of how to keep sites up and running in highly available configurations arose.
It turns out that in single data center environments, you'd have all kinds of equipment; "phantom" routers (a fancy word for "standby") that checked constantly to see if the primary was offline; if so it would self promote.
I call my staging environment "theory" because so many things work in theory but not in production. This is no exception. It turns out that heartbeat failures caused more outages than they solved for in many scenarios.
The "correct" way to do this if possible generally became "have another site you can fail over to." That had the advantage of separate power grids, separate control planes, etc. If the primary site goes down, you can fix it, then fail back.
A lesson you learn pretty quickly is that you want the failover to be automated.
You really want the fail back to be manual. Otherwise you're staring at a "flapping" situation that takes both sites down.
You really want the fail back to be manual. Otherwise you're staring at a "flapping" situation that takes both sites down.
So be sure to avoid that particular fate. That said, this service looks awesome--though at $1800 a month per cluster it's not for the faint of wallet.
Just plan ahead for what failures might look like in your environment.
Just plan ahead for what failures might look like in your environment.
Okay I love this so much. When the @awscloud console asks you to acknowledge that something will cost you money, that's a great indicator that you should check the pricing page.
If you proceed past THIS, you kinda don't get to complain about the bill.
If you proceed past THIS, you kinda don't get to complain about the bill.
"Check these four boxes, because after you do you're going to owe us $36K" is likewise a stellar example of customer obsession / earning trust by @awscloud.
...but I kinda feel like the Savings Plan purchase screen should have a few more warnings than this.
+1 to @awscloud for "yeah, call us and have a chat before we agree to roll a tractor trailer to your office you clown."
It's smart of @awssupport to gate "sign up to pay $180K for Enterprise Support" behind root credentials as well.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
