Question about schedules

Lokk like there is a misanderstanding.

The timeframes you haave in your schedule are for starting the jobs. The jobs dont have to finish in this startwindows.

Your setup is start the backup between Fr 6pm and Mo 6am. is this want you want ? A time windows of 30min is sufficient. If you have a lot of other jobs running occupieng your ressource (tapedrive for ex.) you may have startwindows of 4 or 8 hours. A start windows of 2 and a half day I have never seen before.

I have seen this sort of setup many times .. all jobs kick in at the same time and many remained queued until a resourse is free for it to "run".

If the window closed before it actually changed to "running" then it would fail with a missed backup window.

When there are several hundreds or thousands of jobs running over a weekend it works well to have them sit and queue until a resource is free.

Tunix, we are backing up file servers with tens of TB to tape. Some backup jobs take like 30h (800GB, optimising this is my next step but I asked Veritas and this probably is because we are backing up tons of small files at 3MB/s, and they couldn't give me any advice on how to optimise this)

So those jobs block the drive (I multiplex 3 jobs to tape at the moment) for quite a long time, believe me, full backups of this server in particular will take 40-50 hours easily.

If I leave small execution windows my jobs will fail with error code 196, hence the big execution window.

---

Mark, glad to hear they will run just fine. I'll keep an eye anyway. Do you state they will be ok based on experience or is there any documentation confirming this?

Also, my case is similar to what you describe. I have a robot assigned exclusively to this server, and it has 3 jobs with multiple data streams hence NBU creates a child job for each item in the backup selection. This is how I want it to behave as it allows me to restart specific jobs if they fail, not the whole thing.

That sounds fine .. if you have large servers with lots of smal files then Flashbackup / flashbackup-windows is the way to go .. it costs you an enterprise client license (unless this server is a media server?) but can be worth it as it backs up the drive raw but allows individual file restores.

It does backup the whole disk though so a 4tb disk gives a 4tb backup no matter how full it is .. but it is very fast!

Better would be to use dedupe with accelerator but that need disk as your target.

I find multiplex values up to about 6 are fine

Yeah, I heard about flashbackup but you know how it is when it comes to budget... First I want to have a backup system that works, even if it's not perfect, and later on this year I'll see how could we make it better. Maybe flashbackup, maybe backing up the servers at a ESX level instead.

For next full backup I'll take a look at increasing the multiplexing value.

Thanks again

If you can find the budget to convert to capacity then you can also have dedupe - you could then do vmware backups with accelerator .. you will be amazed how fast that is!!

Good luck!

My previous message seems to be gone?!

Forgot to ask: As I mentioned before, my CINCs run every work day that day at 6pm till 6 am the following day. For instance Monday 6 pm to Tuesday 6 am, and so on. This includes Friday 6pm to Saturday 6 am.

I configured my CINC schedules to not run on the first friday of the month so it doesn't clash with my full backups, but considering the execution window extends to the following day, I'm afraid NBU will start the CINC jobs on Sat at 00.00.

Any way to prevent this from happening for this weekend specifically? Remember that I can't exlude the first saturday too, as first Friday and first Saturday of the month may happen on different weekends. So far I manually excluded the Saturday following the first Friday of the month for the next 3 months, but I wonder if there is another way that doesn't require manually configuring this...

---

Mark, what do you mean by "convert to capacity"?

If the full weekly backup is still running NBU will not fire off the daily inc .. it will wait util the full has finished so there shouldnt be a clash.

If you are on what they call traditional licensing .. so you buy Enterprise Server licenses, client licenses, database licenses, tape drive licesnes etc. then you can convert to what is call a Capacity License

It is a lot simpler as you have just one NetBackup license which is based on Front End TB .. in other words, if you do a single full backup of everything how many TB does that add up to .. that is your FE TB size.

You just buy that license and you can have as many master and media severs as you like, backup as many clients, databases, VMware servers, have as many tape drives as you like and use pretty much every agent and option as long as you keep within that FETB figure.

So Enterprise Client, dedupe and accelerator all come into play.

It can be very expensive so you would need your Veritas Partner or Account Manager to work a price out for you

It is very very important to keep within the FETB figure.  As a backup admin you will be quite within your rights to refuse to add any backup clients which might tip you over your usage.  If you do go over, and Veritas ask for a capacity report, then Veritas are legally and contractually entitled to demand full and immediate payment for all over-usage, and also for back-dated support (if you have support) - at LIST PRICE.

It can be very very expensive for your company/customer/employer if you go over the FETB license usage.

It is quite easy to keep an eye on FETB usage, using the "nbdeployutil" command.  There really are no excuses for using more than you have paid for.

Good working practices of ITIL Capacity Management should save one from any embarrassment.