Web Hosting Talk : Web Hosting Main Forums : Hosting Security and Technology : backup mechanism

[dektong]

In the process of setting up my web server, I have to decide how I would like to do my backup. For sure I will just throw in a second HD to do my backup. Now, what is better (in terms of stability, easiness, etc) to do software RAID or just to mount the second HD as /backup and do cron job backup? BTW, I will be running RH Linux Server 7.0

cheers,

[Chicken]

Well, RAID will really just mirror the drive. It isn't what I'd call back up exactly. This means that if you screw up something, you have another drive with the screw up. If someone cracks the box, the changes they made are on two drives.

In other words, RAID is good in that if the drive fials, there is another one. I'm not sure that's what you want or mean by 'backup'.

[dektong]

Thanks Chicken. yea... I should have known that RAID is not a real backup solution. I guess I was thinking different when I typed that

What I'm really looking is a backup mechanism which will allow you to easily recover the system when the drive fails. RAID allows you to recover easily but it's not actually a backup solution. Having an extra HD as backup (as in http://webhostingtalk.com/showthread.php?threadid=3338) is good but it won't give you the ability to recover easily/instantenously when the primary drive fails. So how do you combine the two functionality here?

I can see it with using 3 HDs, 2 being RAID the third one as backup HD. So when the first HD fails, I have a mirror that will replace the failed HD. Clearly the backup HD will kick in when it is needed.

Anyhow.... If I use only two HDs, one being the primary and the other as backup, then if the primary HD fails, though I have the backup, still I will have down-time to replace the primary HDs (install the OS again, configuring all settings, etc on a new HD). How do you minimize the down time or even no down time (as in RAID)?

cheers,

[kunal]

hmmm.. couldnt something like this be possible... you setup to hard drives. take back up on regular intervals on the other... not as a RAID.. but just access the second when backups are to be taken.. make the 2 like images of each other... so when one fails.. you can just enable the other... i think this is possible.. this would reduce the downtime to 0..

[dektong]

Kunal.... I was thinking about that too. So I assume the way to do it is to install the OS on the two drives (separately but identically) and then to mount the second HD as needed (to do backup). But I am wonderring, when you mount the second HD, will Linux allow you to have 2 root directories? Well, I just can't see why not....

This will reduce down time but it won't eliminate it since when the first HD fails you still need to go inside your box to make the server boot on the second HD. Whereas using three HDs you can practically eliminate your server downtime (due to HD failures).

Anyway, thanks...

cheers,

[cperciva]

If you want to periodically backup from one disk to the other without any downtime, use software RAID.

Tell it that you have a RAID 1 system, but with one drive currently disabled; when you want to backup all your data, tell it that you are replacing a hotswap drive, and it will commence "rebuilding the mirror" -- ie, copying all the files across. Then disable the second drive.

If you just mount the second drive separately and copy files across, then you can run into inconsistancy problems (especially with databases) if files are being written at the same time as they are being copied across. RAID solves this problem by sending all disk writes to the second drive as well.

[scottlaw]

Just remember, if you store your backups on the same server... Even if you unmount the HD, a malicious individual can still destroy all your data. Just look at this thread http://www.webhostingtalk.com/showth...?threadid=5286 If he would not of had tape backups then everything would of been lost.

I know for a fact most hosts do this but I would not be willing to take the chance.

Scott

[cperciva]

Well, that is why you have the backups performed automatically and protected by immutable files, while running the entire system at BSD securelevel 2.

Someone breaking root doesn't have to give them the ability to trash your system if you set it up properly.

[scottlaw]

Cperciva,

Immutable files? I have never herd of them but I am not a system admin... Can you explain a little more about this it should interesting.

Scott

[cperciva]

man chflags

Basically, an immutable file is one which cannot be altered; append-only files can (duh!) only have data appended to them.

These are primarily useful in combination with securelevel>=1 which makes it impossible to remove the immutable and append-only flags.

Generally speaking, a good way to make a system secure is to make all important files immutable and set securelevel=2 during the boot scripts, and when you want to upgrade your system boot off of a floppy (assuming of course that you are the unique person with physical access to the machine) remove the flags, upgrade, replace all the flags, and then reboot back into normal (secure) operation.