I have seen some discussions using interface groups with DDBoost.
I would like to configure it correctly from the start. I understand that I just add the interfaces to the Interface Group in my DDBoost IP Interfaces tab, configure the clients, and DDBoost will find the least used IP and use that IP instead. (If I'm wrong, please correct me!)
I have also seen some people discuss adding a 3rd, LACP device or Failover devices to their Interface Group? If we only intend to use 2 physical connections to this Data Domain using the DDBoost IFGroup, why do I need this 3rd IP and entry in my IFGroup?
For Example, I have configured...
sysadmin@Some_Datadomain# ifgroup show config all
Group-name Status Interface Clients Replication ------------ -------- --------- ------- ----------- default disabled 0 0 0 Site A enabled 2 1 0 ------------ -------- --------- ------- -----------
Group-name Status Interfaces ------------ ------- ------------ Site A enabled 1.2.3.4 Site A enabled 1.2.3.5 ------------ ------- ------------
Group-name Status DD Boost Clients ------------ ------- ---------------- Site A enabled * ------------ ------- ---------------- No replication mtrees with remote hosts in ifgroup File replication is allowed on ifgroup. Client may use any interface.
And my networking looks like this...
sysadmin@Some_DataDomain# net show settings
port enabled state DHCP IP address netmask type additional setting
LACP should only be used if your stuck on NFS/SMB usage. The reason you might create a failover pair is the primary IP for BOOST needs to be up for initial connection. If you loss the primary, the initial connection will fail and so will boost. Once the connection is made the IFgroup is cached and a port/nic/route failure will be handled.
So the best practice for boost is to have the primary NIC setup as a failover pair, and all other nic added to the ifgroup.
So, to make sure I comprehend properly, I would do the following.
Make eth1c and eth1d either a LACP or Failover virtual device, veth0.
Point my applications to that veth0 IP. (Avamar, other DDBoost apps, ect)
Create my IFGROUP with the 2, single port IP address. eth1a and eth1b in my example above.
Watch my apps connect to the DD using the veth0 address initially, but then use the IFGROUP IPs once the DDBoost kicks into gear to send the bulk of the data.
A follow up question to this, if DD replication is setup, would it use IFGroups at all, or would it only use the veth0 port I would set up?
rugby01
85 Posts
3488
0
Posted March 19th, 2018 08:00
LACP should only be used if your stuck on NFS/SMB usage. The reason you might create a failover pair is the primary IP for BOOST needs to be up for initial connection. If you loss the primary, the initial connection will fail and so will boost. Once the connection is made the IFgroup is cached and a port/nic/route failure will be handled.
So the best practice for boost is to have the primary NIC setup as a failover pair, and all other nic added to the ifgroup.