Veritas NetBackup: VMS client

[Michelle Popejoy]

I'm setting up the VMS client for Veritas NetBackup 5.0 MP1.

The manual says that VMS clusters are supported.

For simplicity's sake, lets say I have a two node (FOO & BAR) cluster, going by the common name ALPHA and I have 10 disks DSA0 through DSA9 to be backed up.

Do I...
1. Add one client (ALPHA) and include DSA0 through DSA9 to be backed up,
...OR...
2. Add two clients (FOO & BAR) and assign 5 disks to each?

TIA,
Michelle

[Kris Clippeleyr]

Hi Michelle,

I'm not familiar with Veritas NetBackup, but...
If NetBackup can do more than 1 system at the time, you benefit from defining two clients and spreading the load among them.
However, you lose the benefit of a VMS cluster that way. If one of the nodes goes away (for whatever reason), NetBackup cannot connect and hence can take no backups. I would go for the 1 client solution, just define only 1 client, your cluster.

Greetz,
Kris (aka Qkcl)

[Ian Miller.]

is the common name ALPHA a DECnet cluster alias? Then if the connection used by NetBackup is DECnet then check for the outgoing alias attribute on the session control application in NCL or the object in NCP. If outgoing alias is true then the connection will appear to come from ALPHA and if false from the indivual node. If the connection is TCP/IP then the DECnetg alias is not relevant.

[Michelle Popejoy]

Thanks for your replies.

To answer you, Ian. Sorry, I had the explanation for how we're aliasing the cluster, but removed it when I changed how I phrased my question... It is using TCP/IP. We have a DNS entry for the name ALPHA in with FOO's IP address and another one with BAR's IP.

[Ian Miller.]

I suspect you will have to define two clients as the IP connection will have a source address of the two nodes.

It depends on where the backup is going, if having two jobs will take less elapsed time than one.

[Jan van den Ende]

Michelle,

what I would be most curious about:
if you specify (a) NODE address(es), and that node goes away somehow, how easy can you then restore to another address?

Of course, using the (ip-) cluster alias, that is automatic: if the node goes away, then the address will fail over (if configured ok).

Worth testing BEFORE deploying in production!

Proost.

Have one on me.

jpe

[Michelle Popejoy]

UPDATE...

We *think* we can use the alias (ALPHA). Unfortunately, the development/test environment only has one node. (I know, I know, ideally it would be the same as production for just these situations, but...). Though, come to think of it, we could partially test it by adding two host entries for a new test alias. One with the test machine's IP, the other with a bogus IP. We could at least see if it gets through to the good one.

We're testing some other things (use of logical names containing disks to be backed up, multiple streams, etc.). And when we do go into production there will be a period of time when we're still doing our tried-and-true VMS Backups.

If anyone is interested, I can update as things develop.

[Jan van den Ende]

If anyone is interested, I can update as things develop.


Please do.

Proost.

Have one on me.

jpe

[William Webb_2]

Have you considered renting or borrowing a small node from someone for testing purposes?

[VMS Support]

Michelle,
How are you getting on with Netbackup. We have been running it for approx. Two years. It does have the odd problem but we have overcome them. We have been testing V5 since MP2 and have found various issues. We reported these and they were fixed. This has now got us to V5 MP4.
The main problems were on the restoring of ODS5 type file names.
The latest issue is as detailed belo. This is with Veritas (Symantec) engineering and we are awaiting a fix.

See problem statement below...
I have opened a thread in this conference with reference to this.


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

We run V4.5MP4. We did have various issues with early versions. However this version is more stable. You have to take care when doing restores and the use of *.*.* and *.*;* etc.

We started extensive testing on V5 about six months ago. I found some issues and they were fixed. We then tested against V5.0MP2 and found problems with restores of ODS5 files. Netbackup broke with file names like UPPER_lower_UPPER.DaT ...
Also found some other breaks.
MP3 fixed these issues.
I can provide details on problems and workabouts off line.

Been testing MP3.
Passed all tests.
Next step rollout to test and development.
Run on development servers (4 servers) for One month. All o.k.
Roll out to test cluster.
This is when I found the next problem.

1 GB file - When trying to secure data netbackup fell over with Exquota errors.
Same file on different disk secured fine.
??
Spoke to veritas and they said increase Bytlm on performing backup. This is the System account. Netbackup creates BPCD background process to perform backup.
Bytlm set to 64,000 on dev server that secures 3GB > files with no issues.
Set bytlm to 250,000 and rerun backup of failing file. Worked o.k.

While later fell over again on a different file.

On checking it appears that netbackup is not securing very fragmented files.
The file in question had 25,000+ extents and was only 130,000 blocks in size.

Sometimes you do have files like this.
File is always locked open ...grows and fragments. Very low I/O file ...
The point is that the file should still secure.

Secured file using OpenVMS backup (System account login) and worked ok.
Secured file using old version of Netbackup V4.5MP4 and secured o.k.

Call open with Veritas.