Experiencing a situation where my underlying SQL server is down (DASP working on it, this is not a ticket thread), but it has highlighted that my site throws a completely unhelpful runtime error when the DB pukes. I'm just curious how admins mitigate DB server loss (not data loss). What fall back options are there? Is there a way to have a concurrent backup on DASP and then kick over should the primary server not respond?
Technically, yes. Realistically, I'm afraid not. Mirroring a SQL server - or any server - is an expensive proposition. If you maintain a million dollars worth of hardware (I'm just using that as a simple round number, the real number in our case is higher than that), mirroring those servers requires not only another million dollars worth of hardware, but also twice as much data center space, twice as much electricity (it's easier to get space than electricity in most data centers), increased staff to maintain everything, and not insignificantly, twice the licensing fees to Microsoft. They don't care that a server is a "backup" - if you spin it up, you have to pay them. That's why no one who does shared hosting on discreet servers in our price range does mirroring. It wouldn't be sustainable for the prices we charge. So we backup the databases themselves (which is not a trivial cost in itself), and have procedure in place that allow us to bring up a new server relatively quickly. It isn't a perfect system, because as we all know, all hardware eventually fails. But it's the most efficient and cost effective way to do it. You can certainly find a host somewhere who will provide you with a fault tolerant SQL server, but the cost is going to reflect that.
I can appreciate all that. Cost of service and what not. And I don't think we need the full shebang of Enterprise Hosting just yet. But say I paid for a backup database (on a different server) under your current pricing brackets and then perhaps for the backup/mirroring service once a day... something that just prevented the site from going down, ya know? Big picture, there's a reason I'm with you guys, and its because this rarely happens, but I was just curious what I could do.
A busy site would have an issue with once a day mirroring (if the primary database was unavailable and the "mirror" image was 23 hours old, for example, you'd have data loss). When I mentioned mirroring in the previous post I was referring to live replication. If we decided to do fault tolerant SQL, it would be live replication and it would be an expensive additional service on top of the standard plans. But just to be clear, we don't have any plans to offer that kind of service. But if I've learned anything in this business it's to never say never.
