Good backup procedure for web hosting
How to Keep a Properly Backed-Up Hosting Account
Is your hosting data safe and secure?
Anyone looking for backups is looking to answer a simple question of "What happens if the server I'm on has a catastrophic failure", or in extremely rare cases, "What happens when the entire datacenter erupts in flames?"
And in some cases, "What happens if I make a mistake while editing my website?"
Let's first discuss the difference between redundancy, versus backups:
Redundancy, simply put, is always having (at least) 2 copies, real-time, of the data. This purely solves for technical malfunction of the underlying hardware, as all computer storage hardware tends to wear-and-tear over time. For our part, we have our servers configured in such a way that someone could, quite literally, walk up and pull one of the hard drives out of the server (though we ask that our datacenter techs not attempt this for funsies). If this happens, the server will automatically rebuild the failed harddrive onto a standby drive over the course of approximately an hour, during which time server performance would likely degrade by about 20% efficiency, and after which, it would be able to withstand another hard drive failure, even without the first once being replaced. (For the IT nerds: We use RAID10 SSDs with 1 hot-spare).
The important thing to note about this is this is not a "backup", this is simply meant to mitigate the most likely and common cause of storage failure - but it doesn't protect against a datacenter fire, and it doesn't even protect against software issues like accidental data deletion, because the data on one hard drive is always copied over to the data on another harddrive instantaneously.
Backups are significantly different from redundancy: Backups are what keep you safe from everything from malware to malicious individuals to accidentally spilling coffee on your laptop. (While we're here, a fun fact: While you shouldn't test this theory, pure water contains no metals, and doesn't conduct electricity - so as long as your laptop is dust free, 100% pure water is harmless to a laptop. On the otherhand, sodium/salt is a conductive metal - which is how yours truly discovered many years ago that spilling a cup of ramen broth on a laptop will make for an extremely quick demise.)
Backups are also what keep you safe from a series of hard drive failures, even in a redundancy situation like the above - which, while rare, caused the loss of tens of terabytes of data at another hosting company several years ago.
At CoHosta, we stand by the "3-2-1" rule, with a level of redundancy at each layer. The 3-2-1 rule of backups says that for anything you care about, you need at least 3 copies of the data, on 2 different mediums, with at least 1 of those copies off-site. (Off-site meaning, if you're backing up your personal computer to a USB drive and keeping that USB drive on a shelf next to the computer, this doesn't do you any good if your house burns down.)
Redundancy can (arguably) be one copy of the data. In addition to the redundancy we mentioned above, we maintain 30-day incremental backups on a northeast storage provider (separate from our main datacenter). By only backing up files that have changed, we're able to save a considerable amount of storage space while providing easy access to any recent backup.
For us, however, 30 days of backups didn't feel like enough. We've seen too many cases in which a customer lost all of their data due to allowing an account to lapse, and maybe they didn't notice until months later. (How someone doesn't notice that their website has gone missing for over a month is lost on us, but we have seen it happen, many, many times.)
We've also seen cases in which we wanted to, for one reason or another, recover the exact code behind a website that had been rebuilt with the original version discarded, and then rebuilt again, such that the completely original version was long-since gone, which led us to begin testing a crazy idea:
5 years of hosting backups.
Seriously. Customer or not, our goal is to keep every website customer account backed up twice a month, with those backups being kept for 5 years. We're able to achieve this via a trade-off in which, while the data storage itself is extremely affordable, retrieval is extremely slow, and extremely expensive - resulting in up to 5 days of lead time for individual accounts, or 24 hours to recover an entire server. Referring back to the east-coast incremental backups mentioned above? Our long-term storage is maintained in the northwest: Away from our server datacenters, and on the opposite side of the country from our incremental backups.
The most important backup method
Finally, we want to cover the single most important backup method: Your own backups. No matter who or what it is, the most important backup is the one you hold yourself.
To that end, we enable customer access to JetBackup configuration within our cPanel accounts: Meaning you can sign up for your own storage provider somewhere (SFTP, or S3), and our server will back up your account to that place, too.
Our plan is to have you covered - for a really, really, really long time.