Best way to check integrity?

KGhaleon

I have someone who uses ArcServe to backup the data on their server. They were asking me if I had any ideas on checking to make sure that the files will be restore-able in case of a future problem. I don't know if ArcServe has any built-in integrity checker...but do you guys have any suggestions for this individual looking for some peace of mind?

He's backing up all his data to tapes.

KG

TheShadow

Most people will tell you that the only way to be sure is do a full restore on an alternate resource. Many consider integrity checkers to be the fox guarding the hen house. If the backup has a flaw how do you know that the checker does not especially since tape drives have built in read after write.

Restoring to a spare system or spare disk adds fiduciary validity to the recovery plan. I know of one very large company that fired the entire IT department because grandfather, father and son backups all failed at a crucial time costing loss of many zeros.

Slowhand

Yup. Performing a complete restore is really the only way to ensure that the data is there, the data is intact, and that it can actually be restored when called upon. I remember this coming up a lot during my Security+ studying, talking about ". . .not only backing data up, but being sure that data is intact, is essential to a complete backup solution." Fire up that test system, and restore the data, and do a test like that on a regular basis.

sprkymrk

TheShadow wrote:

Restoring to a spare system or spare disk adds fiduciary validity to the recovery plan.

Fiduciary validity? I say that statement gets my vote for the coolest term of the year posted on TechExams.net! My IQ jumped a couple of points just reading it.
Seriously though Shadow, you are right about restoring being the only real way to test your backups. We occasionaly have to restore lost/deleted files for users, usually a couple of times a month. It's important to know you can rely on your backups. We have had hiccups in the past, but fortunatley it would be isolated to the backup job not running on a particular night, but we were able to use a day-old backup instead. Even that is embarrassing....

Danman32

Backup job failures is one reason Disk to Disk to Tape backup solutions are becoming popular. With DDT, you get a buffer in case of tape failure without a huge risk of losing data. Some shops can't afford to lose even one day of data, so if you had a system failure the day after a failed job the previous night, stuff can go flying.

I remember in my previous job consulting with the IT director over a problem with one of the RAID drives failing. We weren't sure how to replace the drives without killing the array (this was 10 years ago, RAID 5 was somewhat new to most people). He thought we were safe since we had tape backups, until I reminded him that the night backup had not run yet, so the backups were 1 day old, and we had lots of sales transactions that day. He agreed we needed to be extra careful.

JDMurray

Just to be safe, any tape backup should be considered "bad" unless it has been restored to a disk and verified against the original source. I guess this would be a Disk-to-Tape-to-Disk-and-Compare (DTDC) solution, but don't quote me on that description.

Danman32 wrote:

I remember in my previous job consulting with the IT director over a problem with one of the RAID drives failing. We weren't sure how to replace the drives without killing the array (this was 10 years ago, RAID 5 was somewhat new to most people).

This brings up a an important point: Just because a system is "RAID" doesn't mean the drives are hot-swappable. People are fascinated by the hot-swap RAID5 demos where the sales guy pulls a running drive out of an array without crashing the system. The customer then buys a cheaper RAID5 solution, tries the same thing, and destroys the strip set along with their data because the solution they bought doesn't support hot-swappable drives. I've seen it happen.

TheShadow

sprkymrk wrote:

Fiduciary validity? I say that statement gets my vote for the coolest term of the year posted on TechExams.net!

Would you believe that I was drunk? Actually that is a term that you learn if you ever have to testify or support an errors and omissions claim. In this context it literally means legal trust that the backup plan should have worked. "Mr. Smith how do you know that valid backups existed and that they worked"... "Well because the log shows that backups for the last 13 days of the month before the day 14 error were completed. A complete restore to a test system was performed and verified to validate the backup plan less than 90 days ago. Complete restores are performed every 90 days and a dual signature report filed."

sprkymrk

TheShadow wrote:

sprkymrk wrote:

Fiduciary validity? I say that statement gets my vote for the coolest term of the year posted on TechExams.net!

Would you believe that I was drunk? Actually that is a term that you learn if you ever have to testify or support an errors and omissions claim.

Ahh! Lawer talk - that explains it. Good point though.

KGhaleon

jdmurray wrote:

Just to be safe, any tape backup should be considered "bad" unless it has been restored to a disk and verified against the original source. I guess this would be a Disk-to-Tape-to-Disk-and-Compare (DTDC) solution, but don't quote me on that description.

Quoted you.

That's a good point though.
It would consume a bit of time, but I'm sure testing the backups on a regular basis would save myself from a lot of work in the event of a failed drive.

KG