Well this could be a big
discussion.
First Max jobs per client
I totally disagree with you.
I have some servers that are so
small that more than 1 stream causes it to overload, so it does not respond, so
NB kills the jobs.
So I HAVE to set those to 1
stream. My medium size servers I set to 2 and my big ones I set to 4
Reason:
I do mulit streams on all servers.
A big server could have 10 file systems.
So for example I have 1 tape drive
set to 10 streams, when the jobs kick off the first server would try to start
as many streams as it can – so this one would kick of 10 streams going to
1 tape drive. Now 10 stream could really bog down this server to the
point that it cannot do its nightly processing very well. So I now have 1
server beating its brains out to send data to the tape drive ( and I know that
drive will be waiting for data as one server just cannot keep up). Doing
it this way the backup will take longer to finish because the server is slower
trying to send data for 10 different streams
But if I limit how many jobs each
sever does.
ServerA kicks off 4 jobs
ServerB kicks off 1 job
serverC kicks off 2 jobs
ServerD kicks off 2 jobs
ServerE kicks off 1 job
(each server may have more jobs
queues)
Now I have 5 servers doing a bit
more work, and able to do it faster, and still leave cpu avail for nightly
processing. It is my belief that backups run faster this way.
I came to this conclusion when we
first installed netbackup and I had to fight to get the small severs to even
work, and playing with the job streams I got backups to run much better.
1)
Storage Units. I
am not sure I agree with you that when you start using a storage unit that it
grabs all the drives even if it is not using them.
But having only 2 drives I
would create two 1 drive units. But with more drives I have a mix,
I have some that are 6 drives (most of my backups use this one) some that are
one drive (because I have a policy that I want all the jobs to go to 1 tape)
and some that are 2 drives (because I have a SAN media server and the amount of
data will go to two tapes anyway)
3)
VSP I leave it
on. The files that get left behind can be removed if you use the windows
PID killer. I have had to get fancy with some where the server has little
room and I would have to set the size or tell it to always use the E drive. But
ever server could have open files and I want to get them if I can.
From:
veritas-bu-bounces AT mailman.eng.auburn DOT edu
[mailto:veritas-bu-bounces AT mailman.eng.auburn DOT edu] On Behalf Of Johan
Redelinghuys
Sent: Wednesday, November 04, 2009 5:03 AM
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: [Veritas-bu] Some odd Netbackup Questions.
Hi
Been working with Netbackup for a couple of years and you
sort of get set in the way that you do things.
There are curtain parameters that you always set the same,
like “Maximum jobs per client” = 99.
“Allow media overwrite” = I tick all of them
…
So I wondered what and why you set some of the following
settings:
1.
Storage Units:
You have 2 x LTO4 drives. Would you create one stu
with two drives in it or would you create a stu for each drives?
# I’ve found that in some cases it better to
create two stu’s. When a policy starts, it “grabs” the
allocated stu and although it only uses the one drive, it holds on to the other
and the next backup queue’s. With two separate stu’s, the one drive
is always available when one is in use. To make use of the other drive if the
policy should need it, the stu is not made “On demand only” and
Storage Unit in the policy is set to “Any Available”. Also if one
drive goes down, the storage unit can still allocate the working drive.
@ Your thoughts?
2.
Multiplexing:
Maximum streams per drive:
# The rule of thumb is – The drive should never
wait for the host to send or receive data. That said, what would you suggest
the “Maximum streams per drive” should be and why for:
LTO2
LTO3
LTO4
# In my case I usually go for LTO2 = 3 to 5 streams,
LTO3 = 5 to 6 streams and LTO4 = 5 to 8. This depends on the server / OS /
network and a few other factors such as NUMBER_DATA_BUFFERS and
SIZE_DATA_BUFFERS, so it becomes an art to get it just right.
@ Your thoughts?
3.
VSP:
Netbackup creates a VSP directory and creates a *.vsp
file in this directory. In some cases it does not remove this *.vsp file after
the backup completes and eventually it fills up all the space on the
server’s drive. When you try and delete the file, you cannot as it gives
“in use” error. Only way is to reboot the server and delete vsp
file just after the server comes up again. (There is a tech note on how to
delete without reboot, but few clients opt for this option).
# We disable VSP for all clients. This might not be
the best option, but it causes the backup to start off without delay and
prevents the vsp file to be created. We then only allow VSP to be active on
servers that needs to backup open files.
@ How do you guys handle this one?
Some others you guys can think off?