Scheduling Best Practice 1

[Tommojunior]

I'm lucky enough to be the guy who is to configure netbackup in an environment of over 150 servers of various flavours from unix to win2000.

So far, so good. Netbackup is behaving and feeding the Scalar ADIC well.

My question is on best practice for organising schedules for the various policies. I have about 200 policies defined, with varying priorities and varying windows in which to perform the backup. I've read up on priorities, which is nice, but this only seems to priorities between a choice of jobs that are due to start at the same time (if 20 low priority jobs are in progress using my 20 drives, when a high-priority job starts, will it bump off a low-priority job? I don't think so).

=======================================================

So my question, which is a complex one, is how to make the best of the available band-width, available drives and available time.

I can't run all my high-priority jobs first, since the windows available to me on those are later in the night. Some have pre and post scripts which take the applications down and bring them back up when complete.

What I want to say to netbackup is, "Dear Netbackup. Please back up every policy I have, in accordance with the windows allowable for each policy. You can run as many jobs together as you like, as long as you make the best of the drives and always give priority to those that need to finish at a certain time, by running them as soon as possible within the allowable window".

=======================================================

I hope you see what I am getting at here Guys. I don't want to hire a person to sit and watch the backups through the night. We've been there before, pausing/cancelling backups just because we've reached a certain time of night when high-priority jobs need to run, leading to a re-run of the low-priority jobs later in the night (ultimately extending the total backup window for the site, leading to delays in off-siteing the media!).

All Thoughts, Experiances and radical suggestions greatfully Received!

Cheers,

Tommojunior

 

[PGPhantom]

What is your configuration?
Master server version and O/S?
Tape Drives - Robotics, stand alone etc?

 

[Tommojunior]

Nice to have an expert on the case...Thanks...

At present, just a single media server (likely to be broken down in the future). single SAN connected Scalar 10k with 20 LTO drives. 10 or so of the servers (HP) backing up across the SAN, others 100 or gigabit LAN connected.

Using version 4.5GA DataCenter, master is HPUX11 RP5470, 3cpu, 6gig memory (well patched up o/s wise).

Clients are mostly HP (N/L/K/D) and SUNs from Ultra 10 to E10k, not sure on the spec of the NT/Novell stuff (not my bag).

All thoughts welcome!

Tommojunior

 

[PGPhantom]

Expert??

Some starting pointers ...
Use multiplexing - Up to four for incremental backups.
Do not use multiplexing for full backups.
Disable OTM/VSP unless required.
Set your incremental backups to start on the hour and the fulls at a quarter past.
Us ehte priority setting (Looks like you are from your comments).
You are on version 4.5 so you can use drive pools - What you could possible do is to create the pools devices, assign them to pools and for your high priority jobs, have an extra device or two in a separate device pool and only have priority backups use that device pool.

e.g. Say you create two device pools named low and high. You have 20 devices. In "Low" only have 18 devices and in "High" have all 20 but #19 & #20 are at the top - This way there will always be an open drive for high priority backups.

If you have not already done so - Check the FAQ's as I wrote one up to change buffer setting etc - These will make a huge difference in performance. You should be seeing 10,000 - 17,000 KB/sec on each drive. faq776-3124

Hopefuly this will get you started.

 

[kcavanagh]

Not sure I understand why you don't want to use multiplexing for full backups. Anyone care to enlighten me??

TIA!

Kevin Cavanagh

TIA!
Kevin Cavanagh

 

[PGPhantom]

Incremental backups are based on the archive bit so there is a lot of stopping and starting of the tape device as the server is searching for archive bits that are set.

In a full backup - All data is read and then written so the stops and starts are minimized and data is sent as fast as the device can handle it - This assumes that you have a good network.

 

[chainsaw1]

I read your FAQ776-3124. We have lto2 drives, do any of these settings change ?

 

[PGPhantom]

No - The only difference is the tape buffer but the setting that I specified was for LTO and LTO2.