Configure network interface for replication - DD670 & Netbackup

Question

Hi,

I'm reviewing the config of a Netbackup 7.5.0.6 and Data Domain DD670 environment that I inherited last year. We have two sites with a WAN connection, all replication of Netbackup images is controlled by Netbackup media servers - there is no replication configured on Data Domain (On the web gui Replication tab it says "No replication context selected" and on the cli the command "replication show all" returns "This restorer is not configured for replication").

My understanding is that the Netbackup media servers with the DDBoost plugin installed take no active part in the optimised duplication (replication) of images, other than to order it to happen as defined in the Storage Lifecycle Policies configured in Netbackup. They don't actually move the data themselves, Data Domain does it.

I am rationalising the current network setup and looking to team two pairs of 1gb nics: one pair for backup & restore, the other pair for replication traffic. But I don't know how to force replication through a specific virtual interface. I've looked at a doc called "Forcing replication traffic over a specific nic" (https://community.emc.com/docs/DOC-30122) but it seems to refer to replication defined within Data Domain:

Using the IP address:

replication modify rctx://1 connection-host 192.168.6.48

Using the host name:

replication modify rctx://1 connection-host neweth1

How do I force replication through a specific nic when replication is controlled by Netbackup?

Thanks!

ble1 · Answer

By default it will use primary one I believe.  You can fool it by placing on DD1 in hosts data which uses replication IP and point to DD2 name and on DD2 use replication IP of DD1 and DD1 primary name.

stevelaw1 · Answer

Hi, thanks for your reply! But I don't really understand. Are you saying I should add hosts mappings for a replication-only nic? That makes sense, but I still do not know how to make replication data use a specific nic - that is my problem. At the moment everything is going through one IP address - how do I separate replication data out to a new nic?

jbrooksuk · Answer

Hi Steve,

I imagine something like the following may help for you, assuming your replication (edit for clarity: life cycle traffic/path between DDR's) WAN connection between the DDR’s are in a separate subnet with a separate gateway from your management and ingest.

For you, since you plan to team 2 physical interfaces (LACP?) then you would state veth0 for that interface to use.

Maybe set this on both DDR’s in both directions.

Adding a route to a specifc network subnet to go through a specific interface on DD

# route add -net netmask <> gw eth#

Example:

# route add -net 192.168.1.0 netmask 255.255.255.0 gw 192.168.1.2 eth0b

Note: It will add a network route for the 192.168.1.0 network using gateway 192.168.1.2 via eth0b interface on DDR

HTH,

Jonathan

ble1 · Answer

Your data is coming from outside - so it will use interface it should.  On DD level, as you use intergration wih DD, you say replicate DD1 to DD2 (and vice versa let's assume).  If you do on DD1 {replication_IP_DD2     DD2_front-end} and on DD2 {replication_IP_DD1     DD1_front-end} it will result in when DD1 has to send data to DD2 that is using replication IP instead of front-end that you use for backup.  At least that is the guess.  Give it a try and see what happens.

ble1 · Answer

Not sure about NBU, but with NW 1) reads as following: address used corresponds to device path used and defined on backup server.  If this path is defined over VLAN1 associated name - then VLAN1 is used.  If VLAN2, then VLAN2... etc. 2) is something EMC recommended when I had the same question.  Restore will always follow the same rule as 1) will.  Of course, certain muddy water may be applicable when using ifgroups or more complex network topology, but this does not seem to be your case.

stevelaw1 · Answer

Ok thanks for replies people! This is now my understanding, can someone please confirm this is right:

1) DD Boost backup traffic is directed at Data Domain using the host entry in the media server's local HOSTS files. I can point this to any IP address I wish to use - so this is easy to point at the nics reserved for backup traffic.

2) Replication traffic between DD boxes is dependent on entries in the HOSTS Mapping section of the web gui under Hardware\Network\Settings. If I use another hostname to refer to the Data Domain boxes e.g. DDReplicate1 & DDReplicate2, then I can associate that with the ip address of the nics on the subnet I want to use for replication.

Therefore backup and replication traffic are forced onto different subnets.

Is that right? Seems easy put like that

What about restore traffic? It will be requested by a media server (using DDBoost) that won't necessarily be on either of the above subnets for backup or replication. Data Domain will resolve the server's hostname using either DNS lookup or a local HOSTS entry. How do I know for sure it will use the DDBoost-backup nics for the restore?

Data Domain

Configure network interface for replication - DD670 & Netbackup

Was this post helpful?