Engineering Tips

You may need to hard code routes and gateways. If your network is split into several subnets and these all have to use a single firewall or load balancer as their gateway, try the following:

Network Appliance filers in the early days came with with "normal" 512 sector disk drives. But the drive firmware's built-in check summing occasionally caused parity inconsistancy, or data corruption.

So NetApp created their own method of check summing and introduced it in Data ONTAP 6.0. With this new method, every 64th strip on disk was a check sum of the previous 63. Unfortunately, performance suffered terribly, some say up to 30% overhead, until the engineers remembered the mainframe technology of old, where there were 520 sector disk drives with the check summing built right in to each sector. No more 64th strip and the performance was vastly improved.

As a result, the old, "normal" 512 drives became ZCS or Zone Check Sum--named after how the program slips the drive into zones of data stripes and check sum stripes. The new drives with the 520 byte sector became BCS or Block Check Sum drives because the check summing is "built-in".

With 512 or 520, the 18 GB or the 72 GB disks using ZCS (really software check summing) are interchangable. This means you can take the new BCS drives and add them to a ZCS system and they work fine as ZCS disk drives. The drive geometry combined with NetApp's right sizing yields disk drive capacities of 17 and 68 GB respectively.

Unfortunately, there is no comparative drive for 36GB ZCS drives out in the field. After right sizing, the new BCS drives yield only 34GB verses the 34.5GB of the old ZCS disk drives. This means that a customer running 36GB ZCS disks can add space only by adding old ZCS drives or by adding a whole new shelf of BCS disks and creating a new volume.

Question: Is there a way to get NetApp's standard dump command to automatically sequence through tapes in an autoloader? Or do I still have to manually break things up per tape?

Answer: Find out if your autoloader or tape library supports "Stacker Mode" or "Sequential Mode". Spectra Logic 2K TreeFrogs, for instance, support the former. Your manual can explain how to set it.

The autoloader in Stacker or Sequential Mode only knows that if a tape is ejected then it should remove the tape and replace it with the next one in line.

Once the Library is in Stacker or Sequential Mode, then you just need to specify, in the DUMP command, the drive name for each required tape.

EXAMPLE 1

To dump a volume named "test" to tape you could use:

dump 0f urst1h /vol/test

urst1h is the device name on a sample tape library and passes the commands for unload rewind. Run sysconfig -t to determine the equivalent for your own autoloader tape library.

EXAMPLE 2

Now lets say your volume is 150GB and your tapes hold 50GB each uncompressed. You would need 3 tapes. The command to span the three tapes using DUMP would be:

dump 0f urst1h,urst1h,urst1h /vol/test

In this example, the number of tapes is indicated in the command line. If the filer does not complete the dump in the specified number of tapes and needs more, then it will drop you out to the filer prompt with a device open error. This indicates that the tape library is not ready, but that the filer wants to continue the dump.