amityweb wrote:
From my research so far I have come to the conclusion that I wont get a cloud server as per my definition of cloud. It seems they are all scaleable VPSs, so therefore I can accept that. As long as I choose a VPS with RAID to avoid hard drive failures, then thats OK. And backups. We also take secondary backups to our office of websites (file backups and database dumps) using incremental rsync for the past 14 days (keep bandwidth down). The backup restore of Linode is a plus. So I think I am OK to proceed with the server like this.
Even RAID isn't foolproof, but since you've got a good backup strategy, you already know that

Linode has on occasion suffered double-drive failures that took out a RAID array (in a 4-drive RAID-10 array, there is a 50% chance of surviving a two-drive failure, depending on which two drives fail), but that is to be expected when you've got ~2000 host machines. Statistically, some of those disks are going to fail, and odds or that it'll be a double failure in one machine on occasion. Large-scale applications know that disk failures are going to happen, and plan around it. It's interesting to see how large-scale applications handle it. Google just knocks the whole box out of their cloud and replaces it at their leisure. Netflix designed their caching servers to tolerate a bunch of failed drives and just leaves the machine in place until enough drives have failed. BackBlaze builds in enough redundancy that they do a pass once a week to replace all failed drives.
Getting back to your concept of a cloud, the concept *does* exist on an application level. You design the application to be redundant in terms of data (distributed databases, hot failovers, etc) and to tolerate machines randomly dropping out of the cloud. You need to specifically design your application around this, though, and it's not something easy or trivial. Netflix (which is hosted entirely on Amazon's EC2 platform) takes an interesting approach to this: they have a service that runs constantly which randomly kills machines in their cloud (reboots them without warning, I guess?) just to keep them on their toes. This forces them to design their systems to tolerate failure, they can't cheat on it.
amityweb wrote:
So brings me back to my original question! There must be non-linux experts using Linodes. I am familiar with linux and happy to use the shell, no problem setting the server up, installing Virtualmin and a Firewall and configuring it etc. But I dont class myself as a Linux expert. I am sure there will come a time when some issue with Virtualmin/Apache/MySQL/etc. will cause it to go down and I wont know how to fix it.
So therefore what do the non-Linux experts do for support? Anyone know of third party companies that offer this support service? I contacted one over here in the UK but got no response yet, 24 hours later, which does not look good considering I may need emergency response.
One thing I THINK I feel good about, and thats ditching CPanel for Virtualmin. CPanel is great for admin, but seems many issues I get is because of it. Perhaps problems with the upgrade, or all the error emails I get are from Cpanel. I wonder if I will see less issues not having it, making a more stable server. I might be wrong though!
P.S. This forum is not sending me email notifications even though I subscribed to the topic.
Linode intends to eventually offer such a service (they beta tested it a while back), but there was one that is commonly mentioned around here. vpsbuddy, I think it is?