cancel
Showing results for 
Search instead for 
Did you mean: 

VM backups.....Better efficiency?

Toddman214
Level 6

Netbackup 7.5.0.4

Windows 2008 master and media servers

DataDomain backup target, DDOS 5.4.1.1

 

Hi all,

 

In my environment, there are approx 1100 virtual servers. The way I am backing these up, is through multiple vm policies (approximately 12), selected by DatacenterName, VCenterName, and VMFoldername using the VM query, and then staggering the backup schedule kickoff times. My question is, is it better to do it this way, or pretty much have them all under one big policy, and limit number of jobs per policy, etc? Just wondering what others are doing to efficiently back up many vm's. I'm on 7.5.0.4, so not yet using anything like Accelerator for VMware.

 

 

Thank you!

 

 

Todd  

12 REPLIES 12

DG-2005
Level 5

IMO, i would limit the number of backups done via the master and resource limit option.

 

this will help ensure you dont slam any one datastore at any given time.

Walker_Yang1
Level 5
Employee

hi,

It's not a good idea to include all VMs in one policy, which can lead to poor performance.

For more efficient backup, maybe you could refer to the following KB(NetBackup for VMware best practices).

http://www.symantec.com/business/support/index?page=content&id=HOWTO44423

In addition, it's suggested that you upgrade to latest version(the newest release is 7603) and start to use accelerator feature for VMware backup. Here is a article regarding VMware accelerator for your reference.

http://blog.proact.dk/symantec-netbackup-7-6-english/

Thanks

 

BTLOMS
Level 5

I see you already have them running in multiple policies.

 

Are they going to tape or to Disk / AD / PD ?

 

 

Toddman214
Level 6

Thank you. I'm already looking through that article and do have quite a bit of this in place. A specific issue that continues to haunt me are pages of status 196 failures, but I think its just because something is not set efficiently. Of course, I know what a status 196 means, but I'm not exactly sure whats causing it to happen.

 

 

BTLOMS,

We have two new DataDomains in place. I send the VM backups via SLP to my primary DD, and then to the secondary in a neighboring city. Then, I just allow a vault process to run a tape copy.

 

 

BTLOMS
Level 5

the reason I asked the question was because I too did the same , thinking it would be faster to disk. But I found the backups run the fastest , when I send them to tape when we were on 7.5.05

Then again, each environment is different.

Now that we are on 7.6.02, we use accelerated backups to disk which cuts our backup time in half.

Toddman214
Level 6

I know this is a conversation for a different thread, but with Accelerator, can I assume that its placing the track log out on the VM backup host (a media server in my case), and not on each VM client? I dont install the Netbackup agent on each VM, so I think it will have to place the track log on the media server. But, if thats the case, I can carve out a new lun, attach that to the media server, and move the track long there?

 

 

Walker_Yang1
Level 5
Employee

What do you mean by track log in accelerator backup for VMware?

There is the term track log in accelertor backup for File/Folder, which is located on backed up nbu client, when you use accelerator.

Toddman214
Level 6

Walker_Yang, what I meant was, when using Accelerator for VMWare, isnt the track log placed on the VM backup host? In our case, it is one of the media servers? Since we dont install the backup agent on the virtual clients, the track log would not be installed on virtual clients, right?

Scott_123
Level 3

I think Accelerator for VMware relies on VMware's Changed Block Tracking for this instead of track log like Accelerator for file system backups.

This is why a Full with Force Rescan enabled on VMware Accelerator backup sends all of the data, unlike a file system backup with the same where still only changes are sent.

Scott_123
Level 3

and yes, I agree with postings above.

  • Resource Limits are a lot easier to manage than what you're doing with manually breaking this up into multiple policies, max jobs/policy, multiple STUs with different limits, staggering start times, etc., however, you have to use Intelligent Policies to use Resource Limits
  • Accelerator with VMware backup is fast

You said "send the VM backups via SLP to my primary DD, and then to the secondary in a neighboring city. Then, I just allow a vault process to run a tape copy."

Do you use Vault to eject tapes, keep track of offsite slots, etc.?  Otherwise, why not just use SLP to get it to tape as well?

For more information on Resource Limits, see the following

https://www-secure.symantec.com/connect/articles/how-intelligent-intelligent-policy

Walker_Yang1
Level 5
Employee

Hi, Toddman214

There is no track log in accelerator backup for VMware, the track log only occurs in accelerator backup for File/Folder, which means that for the first backup, it generates the track log under NBU client, and for the followed backup, it will update the track log.

In addition, we can't install track log manually.

There are some state files generated every backup, similiar to track log, which located at Backup host and master, every backup will generate 6 files, as follows:

master:

/usr/openv/netbackup/db/snapshot/vm/

backup host:

/usr/openv/netbackup/online_util/fi_cntl

 

Is this clear?

 

Thanks

 

 

 

Toddman214
Level 6

Thank you walker_yang and Scott. I had forgotten that VM backups rely on the change block journal. Regarding the original reason for this post, I think it can be closed out. We have a lot of work to do on our VMware backup environment, as I am now getting multiple pages of status 196 failures. Ive opened a case with Symantec about this, since that issue has come up over and over again with us. I'll work with them on getting that resolved, and running more efficiently. Also, we still have yet to set our VM backups over san, and still are not using accelerator.