Dedicated Server Backup

[bdominick]

This thread started as a trouble ticket, but Matt thought it might be of broader benefit since my questions are essentially generic. I'd love it if others had insights that would help my decision, or make me feel more at ease with it, since money is something of an obstacle at the moment and I can't just take advantage of every option.

I *think* I'm looking to set up an off-server, on-site backup. Don't have a huge amount of data to backup -- essentially less than 6GB to start. I think I only need daily and weekly retention schedules, which would be about 12GB of storage. GH offers this service in 10GB chunks, so I'd have plenty of room to grow on a 20GB package.

Outstanding considerations include:

If I can get an additional HD put in my box for about $15/mo, whereas the off-server package is $20/mo for just 20GB, are the advantages of off-server backup worth it? What are the disadvantages/advantages of each option? The system will FTP a gzipped backup to a remote server (presumably in the same facility), vs. just copying to a second HD. Or should I just be mirroring my primary HD?

I was also wondering about security of the data. A previous system I used employed on-server encryption, transfer over an SSL connection to a remote server, and storage on media that was allegedly physically removed /disconnected after transfer. That would be overkill for my current needs, but I'm wondering what kind of security concerns remote (same-facility), unencrypted backup raises.

Previous questions in the trouble-ticket thread included what gets backed up. The answer is the /home directory (recursively) plus /var/lib/mysql "as well as some other configuration data."

I also previously asked if I could control the restore process from the remote backups, and Matt answered that I would have essentially full control via SSH and less-specific control via WHM.

Hope this helps anyone else who's wondering these things, and I look forward to any and all input.

[Matt]

Here are some answers for those of you that are interested. Keep in mind this thread is geared for those of you with Dedicated servers only. Shared, Reseller, Semi-Dedicated and VPS have different solutions available to them.

Keep in mind these are "Today's" prices and these are subject to change at some point in the future.

Off-Server FTP backup (same datacenter):
$2.00/GB sold in 10 GB increments. This is monthly recurring.

Off-Server Rsync backup (same datacenter):
$2.00/GB sold in 10 GB increments. This is monthly recurring.

On-Server Backups (Same Datacenter, Same Machine, Separate Disk)
The price varies on this based on what size disk you want and if you prefer a one-time fee or a monthly recurring fee. The one-time fee starts at $250 for a $160 GB HDD and the sky is the limit there based on your needs. The monthly fee starts at $15/mo and also goes up based on your needs.

On-Server Mirroring and or Striping - (AKA RAID)
We offer this for those that want it and you will need at least 3 Hard disks for a reliable solution. 2 minimum for the RAID array and one for on server backups to a standalone harddisk. Price varies here and you'd want to select a RAID option when you perform the server build in most cases. If you already have a server chances are you'd need to order a new one so that the proper hardware can be installed.

If I can get an additional HD put in my box for about $15/mo, whereas the off-server package is $20/mo for just 20GB, are the advantages of off-server backup worth it?

Probably not unless you are extremely unlucky and for some reason your machine catches on fire, but in that case, its probably not just your machine on fire, its probably everybody's, and my guess is the fire suppression system is going to give a nice squirt to those machines that did not catch on fire, so you might be waiting a while for components to dry off before you get your data back anyways.

A secondary drive is a very good choice and more affordable.

What are the disadvantages/advantages of each option?

Local
A secondary drive is also more reliable and more secure than an FTP backup. The chances are extremely thin that usernames and passwords would be intercepted over FTP, but it is a plain0text protocol so if your backups have sensitive data, I'd say don't do it.

FTP
The advantage of an FTP backup is isolation from the same physical unit and can be good for some that want that sort of protection, and don't have particularly sensitive data.

Rsync
The advantage of using an Rsync off-server backup is that rsync uses ssh which is a secure data transmission protocol and it is much faster than FTP. The disadvantage is it is expensive to setup unless you want to write the scripts to do it.

RAID
The advantage of a RAID setup is that, should you loose one drive (or ore if you have more than 2 in your array) is that you can sustain a broken unit and still be serving up webpages, email, etc. If you have a hot swap setup you can get a replacement drive in the machine without ever having to power the unit down to replace the drive.

The disadvantage here is that it is expensive, and is not 100% foolproof, and, that if you have corrupt data on one disk, it can be corrupted on both disks. In addition to your RAID, you are going to want some sort of backup like the ones listed above for archival purposes.

All of our shared servers use local backups to a secondary drive. Some of these servers are in RAID + a local drive. Some clients have custom setups and they use Rsync to servers in the same facility for their backups.

Hope that helps. If anyone else has anything to add, please do.

[andychev]

Thanks that has answered a lot of questions i also had. Further to it can you explain how many backups can be stored. For example:

If backups are set to run everynight and every week. so nightly and weekly backups does this mean that there are only ever two backups for a site? Ie every 24 hours the new backup replaces the old backup. And the same for the weekly backups.

OR

Can a chronological backup system be set up? So you may have two weekly backups (if there was a problem you would be able to go back 1 week or two weeks). Same for daily backups would it be possible to have 6? where the oldest would be overwritten. So in total there would be 8 backups per site allowing for restoration from 1,2,3,4,5,6,7,14 days?

Also is it possible to set backups to run throughout a 24hour day instead of at a particular time to reduce server load? Is it worth it?

[bdominick]

Just to update... I've decided to forego the off-server backup, based on Matt's feedback here. It makes a lot more sense. I can't answer Andychev's questions, but I do know that a lot of the server load Matt was worried about for my backup plans had to do with the offsite FTP.

[Matt]

Thanks that has answered a lot of questions i also had.

My pleasure.

Further to it can you explain how many backups can be stored.

Your first example is correct. The nightly is overwritten every day and then a "random" sample is taken from the daily backup to serve as the weekly backup. Same goes for monthly. So depending on what date your account was created, it has an influence on what day your weekly and monthly (if you have monthly) is archived.

Can a chronological backup system be set up? So you may have two weekly backups (if there was a problem you would be able to go back 1 week or two weeks). Same for daily backups would it be possible to have 6? where the oldest would be overwritten. So in total there would be 8 backups per site allowing for restoration from 1,2,3,4,5,6,7,14 days?

Sure but you would have to write the software, we could write it, or you perhaps can find a 3rd party solution for that, but it is not built into the regular backup system and would not be a good option for shared hosting due to load and disk space requirements..... Its probably why its not built into cPanel. That means if you had a 1 GB site on the server it would actually be using up 9GB of space (1 GB in /home and 8 GB in /backup) and there would be some complications there on how to find a large enough backup disk for that sort of backup schedule, and still be affordable. Might work fine for a small site on a dedicated though.

Also is it possible to set backups to run throughout a 24hour day instead of at a particular time to reduce server load? Is it worth it?

Please explain this one I am not conceptualizing the idea.

[andychev]

Please explain this one I am not conceptualizing the idea.

Traditionally i belive backups are done all at once? So the entire server is backed up at say 2am. I have heard it is possible to spread this over a longer period of time, im not entirely sure how this is acheived. Im guessing again it could be some sort of custom script to limit the processor usage that the backup consumes. Thus the backup takes longer but means the server isnt overworked for a period of time?

Im prepared to be called barmy on this one!

[Matt]

Well that's not entirely true. If the backups start at 2 AM then each individual account is backed up, one at a time and if the load exceeds a certain amount, the process sleeps for a bit to let the load go down. Once the loads are acceptable again, the process resumes.

So the spreading out that you describe is automated more than anything.

[andychev]

Well that's not entirely true. If the backups start at 2 AM then each individual account is backed up, one at a time and if the load exceeds a certain amount, the process sleeps for a bit to let the load go down. Once the loads are acceptable again, the process resumes.

So the spreading out that you describe is automated more than anything.

Super, so really im worrying about nothing! Time for a very english cup of tea!

[bdominick]

There are third party backup providers that do what Andychev is talking about. It's actually the industry standard for them, and it's really expensive. I've set up these services before for clients. One place I used that is relatively well-priced and quite reliable (at least it has been for me) is Data Protection Services Online Backup, Remote Online Backup, Secure Online Backup Services

At the end of your first month, their standard service would have on hand 7 dailies, 3 weeklies (they keep the 7th daily as a weekly) and a monthly (the last weekly, I believe), and they collect 12 monthlies across the year. These are full backups, and you customize very precisely what you want backed up. Their software encrypts and compresses the backup package on your server (only you have the pw -- they don't even know it), then transfers it over SSL to their first site, then a second site gets a redundant copy (they're in New Orleans, so this came in handy for us during Katrina, actually). Oh, and you manage the whole thing via client software on one local machine. So I have a program on my laptop right now that securely (and efficiently) manages the backup services for 5 remote Windows servers. I assume this is the same for their Linux offerings.

But this all starts at like $90/month, for like a Gig, and goes up, sharply, from there, depending on how much data you need backed up. The funny thing is, they only charge by the amount of data you're backing up, not by the total storage you're using on their end, which makes a big difference. Anyway, if I had the money, I'd definitely go for something like that, but that was when I was dealing with clients who had highly sensitive data and pretty substantial amounts of it, at that, like e-commerce sites -- and I was spending THEIR money.

It would be REALLY nice to have a few extra dailies, though -- in most cases, I've needed backups because a problem is discovered a couple or three days after it occurs. Going back a whole week would sometimes be really daunting. But such are the perils of poverty. Will there really be no way for me to store more than one daily on my second drive? I would actually rather have dailies than weeklies, if I had to pick...

[Matt]

Guess we need to get into the backup business.