Pretty High Availability (PHA)- a compromise between High Availability and No Availability
It seems that Cloudron currently targets customers who probably wouldn't generally be willing to pay for High Availability (HA). But I suppose they might be willing to pay one or two US dollars a month for what I think of as "Pretty High Availability" (PHA).
In this case, instead of HA—five-nines of availability (99.999% availability)—how about how aiming for two-nines of availability (99%)? Of course 99.999% is better than 99%. I know it is. But, those extra three-nines are probably too expensive for the overwhelming majority of customers Cloudron currently targets.
Let's imagine a Cloudron user named David has an account at Digital Ocean and another account at Hetzner. Let's suppose that David doesn't want to pay to have two VPSes on 24/7/365. Normally he only wants to pay for one VPS to be on 24/7/365. (After all, David needs money for important things... like beer and more beer). Ok. Fine. Let's say David rents (buys) a VPS at Hetzner and has his data backed up every 6 hours to Backblaze.
Now, let's say David has set up monitoring that runs on Cloudron to monitor his VPS at Hetzner. And then yes.... oh Noooo! The dreaded day finally arrives! His VPS at Hetzner goes down for 15 minutes! Not only does the world stop rotating on its axis but far, far, far more importantly two of his 179 customers who pay him $5 per year for access to his world-famous website, https://DavidsMinecraftTipsAndTricks.com are unable to access the site!
Surely you are thinking something like: David stands to lose millions of dollars in future revenue and these two poor souls who had been on his website will be scarred for life!
But wait! Just hold on. Ok? See, Cloudron has a solution! It's an imperfect solution. But it's a solution nonetheless. While David is sitting on a beautiful, warm, white sand, tropical beach staring out at the crystal blue water, with a beer in nestled in one hand and a beautiful young woman cozily sitting in his lap, blissfully unaware of the tragedy occurring... a rescue plan has already begun to be automatically implemented.
See, a script running on Cloudron will have detected the 15 minute outage on David's VPS at Hetzner. That script will then create a new VPS at Digital Ocean, update the DNS record, and voila... David will have, unbeknownst to him, reap the benefits of PHA (Pretty High Availability )!
Yes. I realize that updating the DNS record can take up to 72 hours. But perhaps something similar to what is explained in Secondary DNS - Deep Dive, and Configuring Secondary DNS might help to cut that time down significantly.
Essentially I am imaging an inexpensive means of moving away from servers as pets and towards servers as cattle. Please see, The History of Pets vs Cattle and How to Use the Analogy Properly.
robi last edited by
@robi Thanks for letting me know.