Operating System - OpenVMS
1748223 Members
4876 Online
108759 Solutions
New Discussion юеВ

Using Backup/Encrypt - Saveset Corruption with no warning.

 
SOLVED
Go to solution
Shriniketan Bhagwat
Trusted Contributor

Re: Using Backup/Encrypt - Saveset Corruption with no warning.

Hi,

/CRC qualifier instructs BACKUP to calculate and include a cyclic redundancy check (CRC) value for each block of data written to the backup file. RESTORE, upon seeing the presence of this check, re-calculates the CRC value and ensures that the new value matches the one written to the backup file. If the two values do not match, RESTORE reports a CRC error and then attempts to recover the block using standard XOR recovery methods. CRC has a significantly higher probability of detecting errors than the checks normally used on tapes and disks. Additionally, CRC detects any corruption which may occur as the data is transferred to or from the computer's memory. There is significant overhead involved in calculating the CRC. Using the /CRC qualifier may degrade the performance of BACKUP.

It is good to use both /CRC and /GROUP qualifier with BACKUP.

Regards,
Ketan
Hoff
Honored Contributor

Re: Using Backup/Encrypt - Saveset Corruption with no warning.

LTO is rated at less than 10^17 bits read uncorrectable, and less than 10^27 bits read undetectable, and incorporates redundant writes and read-after-write support for automatically detecting write errors, and implements data redundancy.

The ECC schemes used are (240,234,7) and (64, 54, 11) Reed-Solomon, which are pretty good ECC.

CRCs are implemented to detect any compression errors, as well.

And yes, you're expecting BACKUP to use and to recover from what is usually compressed data, just to keep things truly interesting.

And for comparison purposes, there are empirical studies showing three to six hard errors per terabytes on hard disk drives. The approach used there is RAID. That's a rate of 3 to 6 hard errors found in 10^12 bits of disk.

Somewhat playing the devil's advocate, I'm not entirely convinced that the BACKUP implementations of CRC and XOR are effective (or even needed) with modern storage, and that writing duplicate tapes might not be the better overall approach for the likely failures.

Put another way, there's the ubiquitous belt-and-suspenders approach that is sometimes blindly applied by various sites, but there's also the "are you even going to get anything back from the tape to even try a BACKUP CRC/XOR recovery?" question? Sure, you might catch a bad device or bad tape, but are you even going to get data back from that widget, presuming the read-after-write doesn't catch the error, or the storage media degrades or later becomes damaged.

I've not seen a study in this area.

Put another way, just because we've always used a particular approach or solution doesn't necessarily mean it's still required.
The Brit
Honored Contributor

Re: Using Backup/Encrypt - Saveset Corruption with no warning.

Thanks for the discussion guys,

Normally when I apply an update patch, I compare the contents with the master list and manually add any which are not included in the update patch, to my list of patches to apply. I dont know how I missed this patch (BACKUP V0100). But I will take care of it since it doesn't require a reboot.

I was not able to verify the backup (on a different system) because I re-inited the tape to create the replacement backup. (this was less complex at our location that modifying the cataloging system to change the tape barcode).

Thanks again

Dave