Your premium source for custom modification services for phpBB

HomeForumsBlogMOD ManagerFAQSearchRegisterLogin

Comments October 8, 2008

Backing Up Your Board…. to Amazon?

Filed under: Backups, Board Management, phpBB — Dave Rathbun @ 11:38 am CommentsComments (6) 

One of the most critical steps that many board owners fail to take is designing and implementing an effective backup strategy. In my opinion, backups cannot be a manual process. They have to be automatic. If they’re not automatic, then they don’t get done… and a backup that isn’t done is worse than no backup at all. At least if you know you’re not doing backups you don’t have anything to complain about. :)

I lease a dedicated server from a hosting company so I have root access. I do several layers of backups, including an automatic download to a secondary server located in my home office… and storage on a RAID (mirror) device. This all runs during the wee hours of the morning. But what if you don’t have the same choices as me… does that mean you can’t have the same advantages?

The other day I noticed an ad being displayed on the bottom of one of my boards via the Google Adsense program. I can’t click on the ad (it’s a violation of Google’s Terms of Service to do so) but they provide a tool that lets me determine what the target URL is. It was for a service called Zmanda Open Source Backup. The “open source” part caught my eye, so I went to their site. In a nutshell, they provide a backup program. That, in and of itself, probably isn’t that interesting. But what is interesting is that one of their target devices can be Amazon.com’s S3 service. Amazon.com offers a data center service where you pay for the space that you use. It’s fully redundant / fault tolerant and comes with a 99.9% uptime guarantee. I would rather see five 9’s instead of three, but that’s not bad.

As a brief aside, for those who are not familiar with the term “five 9’s” as I have used it here. The phrase “five 9’s” refers to 99.999% uptime. To put that into perspective, three 9’s implies up to nearly 9 hours of downtime in the course of a year. Five 9’s implies only about six minutes of downtime during an entire year. Big difference? Only if you need your data at some point during those nine hours. :)

When I reviewed their service description I saw that instead of selecting a tape or disk device as the destination for my backup, I would be able to select the Amazon S3 data center. My own backup process includes keeping a rolling archive of files for each database on my server, including clients that I host. The most recent backup file is kept on the server itself, on a secondary hard disk. I keep generations of backup files on my home network going back quite a while.

With Zmanda’s backup system it seems that I could do the same thing with the advantage that I could run a restore from anywhere with Internet access. Today if I need to restore I have to be at home.

There is a cost, of course. The S3 service from Amazon has a charge associated with it. But I paid over $1,000 for the RAID device that I currently use for my backup destination. I also have to maintain a linux box that runs the cron jobs required to download the files after they appear on my production server. I haven’t investigated the cost of the S3 service but I bet it would be cheaper than that.

For now I’ve already made the investment, so I will continue to use my existing process.

Coming Attractions
I have a few other posts related to making board backups coming out over the next few months (as I get time to complete them). I wanted to start with this one because I thought it was an interesting service. In my next post I plan to detail why backups are important, and how to go about determining a backup strategy that works for your board.


  1. Yep, backups are definitely important. The most number of days I’ve lost are about 8. Most recently, last November I accidentally deleted my site’s main database. I was fortunate to have a backup just 4 days old. Yes, 4 days is quite awhile ago, but I bet there are many other “small time” admins whose latest backup may have been a month or older. Nowadays I make sure to back up the main database at least 2 to 3 times a week, and I back up the other two databases about once a month since they are less important. I don’t have anything automatic in place yet.

    Comment by Dog Cow — October 8, 2008 @ 11:44 am

  2. I used to backup weekly, on Saturday. That was fine until I ran an update statement and set every user to the exact same password. On Thursday. :shock: I shut the board down and restored the database back to Saturday, and therefore lost a few days of posts and new users. Fortunately this was in the early days and it wasn’t that much activity.

    Now I backup every night.

    I had one client who had me set up his script to run every hour, which is a bit extreme if you ask me. He should have looked at replication rather than hourly hot backups, but he didn’t want to listen to me. :roll:

    Comment by Dave Rathbun — October 8, 2008 @ 1:21 pm

  3. To be clear, over the years, the days I’ve lost totals to 8 from about 3 separate data losses. I’ve never lost 8 straight days all at once. In case that was unclear….

    But yeah, it’s a shame. From that November incident, I lost a guy who had signed up, was pretty excited about the site and had even sent me a PM or two. All gone. He didn’t sign up again, so I’ve no idea what happened to him. :(

    As usual, I’m looking forward to your next article on this topic. :)

    Comment by Dog Cow — October 8, 2008 @ 4:19 pm

  4. Backups are great, but do you know how to restore those backups? That is even better. ;) I’ve talked to people that do backups, but haven’t a clue how to restore them properly.

    As for 5 9’s … well, yes, it makes a big difference, but very few can live up to that, including Amazon. Many startups lost customers/data when Amazon’s cloud computing blew up in their face recently.

    Comment by Micheal — October 8, 2008 @ 5:27 pm

  5. I would be interested in hearing more, Dave.

    I do automated nightly backups of my most critical data using mysqldump.
    I do weekly (sometimes more frequent) backups of my entire database using mysqlhotcopy.
    Database dumps are gzipped and I download them to my home computer. The weekly is manual at the moment as it requires me to disable my site for a period of 10-15 minutes.

    My server is RAID 10 but I have been thinking I should probably have another server drive just for online backups. I currently download backups to my home computer, which is also backed up to an external drive, but if something fails online, it takes a long time to upload a restore.

    I do know how to restore. :)

    Comment by Everett — October 8, 2008 @ 7:26 pm

  6. Micheal, you stole a bit of my thunder. :) Testing the backups by restoring to a different system and ensuring the validity of your backup strategy is definitely going to be a post in this series. A very long time ago I used to work for a backup software company. You would be amazed how many people blindly trusted their backup process without ever testing it. That’s why I say that doing a backup without testing is worse than no backup at all.

    Everett: I used to take my board offline during the backup process but I decided that the amount of time it took to do the backup was not worth the inconvenience. Sure I run the risk of some data consistency issues but the risk of that is slim in my opinion.

    My backup takes only a few minutes. The only reason it even takes that long is because of those darn search tables. :lol: Yet the extra minute or two it takes to get the full backup is far faster than the five hours it takes to rebuild the search index.

    Comment by Dave Rathbun — October 8, 2008 @ 11:10 pm

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

Powered by WordPress