Configure media server with two IP addresses?

I dont't think that configuration will work because both NIC are connected to the same IP subnet. The OS routing function will find the first NIC connected to the subnet and use that for traffic. If the NIC connection was on different subnets -(e.g 192.168.1.1 and 10.0.0.2)  it would work.

But honestly I think you are better off using bonding / LACP type 4. It does require some addition configuration and has prerequisites for the networking gear, but allow you go faster than 2 GB and allow link failures occour without down time.

Secondly - it will allow a more simpler setup because a bonded NIC is seen as a single connection.

100% agree with Nicolai, use bonding if you want the same traffic to use both NICs

This is beyond the scope of netbackup, don't create complicated configuration to accommodate something that the OS/Network should handle. As mentioned above, bonding, LCAP, or teaming is what you need.

Thanks for replies! But I can't team the nics if they're on different switches - our antique switches don't support it. And if I patch both nics into the same switch I lose half the throughput as there's a 1gb max on each switch. I really need to utilise both because a 1gb stream to tape is way too slow and we can't keep up.

Nicolai wrote:

The OS routing function will find the first NIC connected to the subnet and use that for traffic. If the NIC connection was on different subnets -(e.g 192.168.1.1 and 10.0.0.2)  it would work.

Hmm, but if the SLPs for different jobs stipulate for the job to use either server1_a or server1_b media servers, which refer to the same physical host but different IP addresses & nics, won't that force the read (this server will mostly just be pulling data from Data Domain and writing to tape) to use either one nic or the other, and thus use both nics and double my throughput?

Remind me again why you're not connecting to the DD via FC?

I don't think this will work as it would figure out you're trying to add the same storage server to the same media server twice.

Create nbdevconfig entries on the media server for both hostnames:
nbdevconfig -creatests -stype DataDomain -storage_server dd01 -media_server server1_a
nbdevconfig -creatests -stype DataDomain -storage_server dd01 -media_server server1_b

Riaan got two very good points

I  think your assumptions are wrong Steve, a OS decides what to connect to, not what to connect from.

Some application can however set a FROM address, Netbackup does when you use REQUIRED_INTERFACE

Do you use BOOST ?

In that case you may be able to use the BOOST link aggregation layer

https://www.emc.com/collateral/software/white-papers/h7296-data-domain-boost-openstorage-wp.pdf

page 13. The documentation is not very well written in my view. There are too many "open qustions" by the end of that section.

Problem here is you have one server, not many as show in figure 6-8 in above document.

Hi Riaan,

"Remind me again why you're not connecting to the DD via FC?"

I do have an FC connection configured from this media server to the local DD box. However all our backups from various servers are written to Data Domain via IP over LAN to a pre-existing disk pool, and the FC connection cannot read from that disk pool - it can only read from a dedicated FC disk pool. The pre-existing disk pool seems to be IP only. The command to register the media server over FC with Data Domain goes like this:

nbdevconfig -creatests -stype DataDomain -storage_server DFC-DD01 -media_server swnb03

nbdevconfig -creatests -stype DataDomain -storage_server dd01 -media_server server1_a

DFC-DD01 and DD01 then act as if they are two entirely separate devices. If I only have an FC connection configured on the media server and try to write data from the IP disk pool (where all the backup images are) to tape, then the job fails because the FC-connected media server cannot access the IP pool.
As a workaround I configured an IP connection for the media server and added an extra step in my SLP to duplicate the images from the IP disk pool, which is "MBFS_BAS", to the FC disk pool "MBFS_BAS_FC", and that works but it takes AGES to run because it is not an optimised duplication like when you duplicate backup images directly between one DD box and another - what actually happens is that the media server pulls all the data from the IP pool down to itself (via IP over LAN) then writes the data back to the FC pool on the same DD box via FC.

So the problem is two-fold:

1) A media server with an FC connection to Data Domain cannot read directly from pre-existing "IP" disk pools (where all the backup data is).

2) Netbackup cannot initiate an optimised duplication within the DD box from the IP pool to the FC pool, as a workaround. The duplication is managed as if the two disk pools are on separate devices - data goes to the media server via IP and back again via FC - meaning this extra stage adds a large overhead to the process and is by far the longest stage.  

This problem has driven me crazy - EMC have no fix except to manually perform a fastcopy 'snapshot' command on Data Domain, which IS an internal optimised duplication. But then the process is longer automatic, and Netbackup won't know that this has happend and will lose track of where the data is.

If you have a fix for this then I will literally SEND YOU MONEY. (It won't be much, but probably enough to get blind drunk on). I will also eternally sing your praises from one end of the internet to the other!

hey Steve

Are we back to this thread here : https://www.veritas.com/community/forums/data-domain-reading-data-fc-and-ip-written-storage-units

Hi Steve,

I remember now, thanks for the refresh. And yes, you're screwed unfortunately. I'll summarise (for my own clarity)

• You've got multiple media servers that write across LAN to DD.
• You have one media server with a FC connection to DD. This media server is also connect to tape. This media server has credentials for 2 storage servers, the IP version and the FC version.
• The FC and IP version are isolated in NetBackup and the fact that they're in the same box is irrelevant. So you basically have two pools on the same DD FS.
• The media server with the FC and Ethernet connection is therefore going to read across IP and duplicate it to tape (across FC).
• Is there anything written to the FC pool? That might duplicate to tape easily.

Since the LAN is the delimiting factor, is it not possible to add the multiple media servers to the SAN so they can join the FC party?