Hello folks,
Does anyone know how (or know why I wouldn't want to) to use the -L option to the dump command when using Legato NDMP. The -L is supposed to do a backup via file system as opposed to inode table. This would seem to me to make restores (esp. of directories) alot quicker. I'm sure there would be a significant hit on the backups, but maybe tolerable. I haven't seen any mention in Legato or documentation of -L in NetApp docs.
Let me clarify a little bit about the 'L' option.
The 'L' option makes NO difference on what ends up on the tape. Therefore, the 'L' option makes no difference on restore speeds.
On toasters, the other day, I talked about how dump makes two bitmaps - one that describes what inodes are in use active in the subtree, and one that describes what inodes are on the tape.
Pass I and II of dump construct those bitmaps.
If you are running a level 0 dump of a volume or qtree, we can figure out those bitmaps based on the inode file. We are accessing one file, instead of a bunch of directories.
If you are running an incremental dump, OR a dump of just a subtree, we run through the directories.
In general, being able to scan one file to determine what needs to be dumped is a huge performance win. However, sometimes if you have a small qtree, it would actually be faster to just look at the directories in the qtree than to scan the entire inode file (which is full of inodes for other trees...)
For those cases, and those cases alone, you might use the 'L' option to see if Pass I/II run faster.
Furthermore, if you are running 6.0 or later, the inodefile walk code was optimized by Steve Fong, so that you probably never want to use the 'L' option.
Like I say, though, the data that ends up on tape is EXACTLY the same, with or without the 'L' option.
Stephen Manley DAM and NDMP Best Actor Nominee