On Tue, 2006-03-21 at 10:23 +0530, Srinivas Rao wrote: 
> 1. What is the logic behind incremental / differential backups?
An incremental backup will backup all files since the last backup,
regaurdless of the level of that last backup.

A differential backup will backup all files since the last backup of a
lower level.  For example if you run a level 9 diff backup it will
backup all the files that have changed since you ran a level full or
level 1-8.

>  How does the Networker know whether or not the file has backedup earlier
>  and with what level?

It looks at the change time of the file and compares that to the date of
the last backup and the level.  It is important to note that NetWorker
does not honor the archive bit, although I believe it will reset it on
filesystems that support it.

> 2. Does it check each and every file to prepare the list of items to be
>  backedup with incremental or differential backup?

Yes, run save with the '-vvv' option and you will see it build the file
list.  What it does is as it walks the tree, it compares the ctime to
that of the last backup.  Then based on the level of the current and
previous backup, either selects the file to be saved or skips it.

> 3. One of our backup engineer aborted inventory operation in progress and
> because of this Networker did not pick up tapes for scheduled backups and
> it did not allow us to do any other operation(labeling etc..due to message
> "Inventory operation already in progress"). However I could see media
> pending requests in the monitor window. Why did it happen?

Aborted how?  Did they make sure that the nsrjb command was killed, or
just use NMC/'other retarded utility' to "stop" the inventory.

Most likely the nsrjb command that the GUI fired off was still running.
Its important to note that NetWorker works through the firing off of
many different commands.  The GUI's just wrap this into a shiny point
and click for the PHB's to look at.

I would also assume you only have one tape drive, as others should have
picked up the slack for the one doing the inventory if they were there,
but that all depends on your config.

> 4. Why index is always being written with level 9?

This is due to a incr or differential backup being run.  Indexes of full
backups will be written with level full.

> 5. Why does Networker write indexes from the Networker server to tape
>  media after backing up the client?

Because only then is the index complete and able to be written.  You
really don't want your indexes all fragmented on tape.  So since it
can't know what is coming down the pipe, it builds the index as the
datastream is happening, once the backup is finished, only then can it
write the full index to tape all in one shot.

> Why is Networker slow with restorations??

Use linux/solaris for your networker servers, period. ;)

Assuming you are already using a Unix like NetWorker environment,
restore time is highly dependant on the filesystem of your client, and
its ability to open and close file descriptors rapidly.

If your running linux on your clients, don't use ext3, its a decent
filesystem for general use, but you'll get better performance out of an
alternative filesystem like xfs, jfs, or reiser.  If you must use ext3,
then make sure you tune it accordingly to the type of data you are
storing.  If your running Windows/NTFS, run away.  Its performance, is
dreadful.

The type of files you restore also have bering on the speed.  A folder
that contains 1G of 1k gif images will take around 1-3 hours to restore.
A 1G file that is a solid file, will only take a matter of minutes.  Its
all about the data stream.  With small files, the data stream starts,
streams, stops, filesystem flushes, starts, streams, stop, flush, start,
stream, stop, flush, etc.  Large files start, stream, stream, stream,
stream, stream, stop, flush.  The filesystem and the OS have to deal
with much less overhead.

Another thing to consider is when/how you do your backups.  Since
writing to tape is a linear operation, data must be multiplexed into a
single data stream.  When many backups are happening all at once, your
data stream becomes fragmented, therefore on the restore end, you will
spend more time seeking on the tape as it demultiplexes the data, than
if you had a single constant data stream.

Just as a sugestion, I would spend a night, with a good strong cup of
coffee, and a pack of cigarettes (you'll need them ;), and read all the
NetWorker man pages starting with nsr(1) following through all the "See
Also"'s.  Contrary to popular belief NetWorker is extremly well
documented, and the man pages provide valuable insight into the nitty
gritty of NetWorker.  Although some of them do need some updating and
better documentation, aghem like the ebnf's scatter throughout being
incomlete, they are thousands times better than the 'Admin Guide'.

Have a good one!

Jason

To sign off this list, send email to listserv AT listserv.temple DOT edu and 
type "signoff networker" in the
body of the email. Please write to networker-request AT listserv.temple DOT edu 
if you have any problems
wit this list. You can access the archives at 
http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER