Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

We use Symantec backup exec 12. We don't feel it is necessary to backup the system state nightly. We are opting to back it up monthly or if system changes are made.

I have read articles on what exactly the system state back's up. From my understanding so far, even with the system state, one still needs to backup their whole C: drive and any other drive on the system to be able to perform a system restore.

We are backing up multiple servers to a compress 800gb tape. The question is, now we opted to only backup up the system state monthly, which saves us 80gb's per full backup, is it your opinion do we need to continue to backup the full C: drive nightly?

Note: We do a Weekly master and nightly incremental. The reason this question arises is that we hold out the weekly master, then a monthly returning the previous week into rotation, hold monthly until the year end is performed and then year end is saved for good and each month will eventually go back into rotation. Blah Blah I am rambling. Currently we are on two tapes per full backup. We were trying to get down to 1 tape, for convenience, but want to do the backups RIGHT!

What have other admins found as best practice?

share|improve this question

6 Answers 6

up vote 1 down vote accepted

Keep in mind that the System State includes the Registry. Whether anything changes the registry will depend on your server and the services installed. If any of those service definitions reside in the registry and the settings change, you will need to update the System State backup after any of those service definitions changes.

share|improve this answer

In theory you could backup strictly the system state and you would be able to restore that to a fresh installation of the OS on new hardware if you had a disaster.

In practice, things don't always go so smoothly. Oftentimes there are either A) logs or B) hardware-specific utilities/drivers sitting on the C-Drive which would be either helpful or critical in the event of a restore.

In order to figure out whether you can get away with just the system state, you should really perform a disaster recovery test and do an actual restoration (in fact, you should really do this test just as a general rule every year). Always err on the side of backing up too much, until you've proven that you can do with less.

-j

share|improve this answer
1  
+1 for disaster recovery testing. There are always some hiccups the first time around for most scenarios; a rehearsal every now and again can make a real DR event much less painful. –  nedm Jun 4 '09 at 22:49

I do complete backups of our C: drives on select windows servers with system states weekly. We only bother to backup the OSes of machines which have things kept on the C: drive that aren't easily recreated. Many boxes are faster to rebuild than restore anyhow. The theory behind Weekly Fulls being stuff shouldn't be changing on those drives all that often. We keep those backups two weeks, so we have two back copies. It has given us a great balance at being able to get back what we need and not consuming too much disk. Also it means our OS/System State backups are relatively fresh, and if we do have to roll back in changes lost, they're only from the past week, which means they are recent enough to remember, and not too many as to be a nuisance.

We do not store user data of any sort on our C: drives. They're jsut there for application data that can't be installed on another drive, and for configuration issues.

We do have nightly backup of our DCs with SystemStates so that we have current AD info once again those we only keep two weeks, basically for DR only. In comparison to user data which we keep much longer.

share|improve this answer

I'm not certain what your situation is, but in my experience a full backup of a system is not necessary or even a good idea. (GASP!)

Essentially, the best thing to do is have failsafe redundancy in the beginning (such as Master-Master replication on a mysql server), or Raid-1. (or 5 for best results).

Basically, in my opinion and experience a "state-backup" is a messy way to get things done. If you need absolute minimum downtime (as in less than a minute) use redundency. for all else, use a backup solution that fits the system (rsync based solution for the file system, repositories for code, etc).

Also, ALWAYS CHECK YOUR BACKUPS. In my previous post as a Jr. Linux Sysadmin, I came across 8 years worth of weekly backups...ended up needing one. found that the man who had just been fired for gross-incompetence, had been backing up the wrong directly (an empty one) for 8 years.

share|improve this answer
1  
RAID is not a substitute to backup... Neither is redundancy. Hardware issues are real, but you should also be able to restore data of that important spreadsheet that was overwritten yesterday and that both RAID-1 drive or the redundant server have the new "wrong" copy. –  Hapkido Jun 5 '09 at 0:51
    
I think he's talking about redundancy vs. state backup, not data backup -- I don't read that he's proposing RAID as an alternative to (or substitute for) backup at all. – –  nedm Jun 5 '09 at 1:07
    
@Hapkido: nedm is right. I'm not saying you shouldn't have some kind of backup, I'm just saying "state backup" isn't something (in my experience) that works out very well. I personally use something similar to time machine to store differences in files for myself. I was just suggesting hardware/redundancy as what I consider a better alternative to state-backups. –  Tedd Johnson Jun 5 '09 at 22:17

We don't do full backups of C:, either, but we do back up system state nightly. If you use active directory, it's pretty important to keep regular system state backups from your DCs -- they don't take up that much space by themselves compared to a full backup of the entire drive, and if your backup is older than the DNS tombstone age (which is 60 days by default) it's no good for restoring AD.

From direct experience, bare metal backup of the OS and system drive often don't go smoothly. We choose to use things like hot-swap RAID for redundancy and then backup data and -- wherever possible -- configuration with the assumption we'll need to reinstall OS and apps if anything goes completely bonkers anyway.

share|improve this answer

Complete backups of asolutely everything on the server here; full at weekends, differentials nightly. Considering the fairly minimal overhead that doing System State imposes on your backup in terms of both time and storage, there's no real reason not to do it.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.