What if the e0M is connected to the same subnet as data?

Sometime you can have problems like:

Network performance over ethernet networks after reboot/maintenance has degraded.
Slow backup and transfer issues are reported on a storage system when performing the following backup and transfers:

  • NDMP backup
  • SnapMirror transfer
  • SnapVault transfer

Messages and EMS logs will warn with the following messages (IPv4 or/and IPv6 accordingly):

[netapp: netmon_main: net.if.mgmt.sameSubnet:warning]: ifconfig: IP address 'x.x.x.x' configured on dedicated management port 'e0M' is on the same subnet as IP address 'x.x.x.x' configured on data port Prod1. Management IP addresses must be on dedicated management subnets.

[net.if.mgmt.defaultGateway:warning] route: Static or default route with gateway ‘x.x.x.x’ is targeted to dedicated management interface ‘e0M’. Data traffic using this route might be throttled due to low bandwidth, or dropped if a protocol filter is configured.
The reason:

Having the network management interface, which is currently a 100 megabit connection, configured with an address that is on the same network as a gigabit, or even 10 gigabit network interfaces, can cause the gateway for that network to get associated with the management interface. This can have a large impact on the performance of the network traffic.

This impact is more noticeable when the e0M interface is configured with an IP address in the same subnet as the default gateway for that system.

NetApp storage systems use a feature called Fastpath to decrease the overhead needed to lookup routes in the routing table by using interface-to-mac caching. If the e0M interface is in the same subnet as production traffic, it is possible that when traffic is received on a 1Gbps or 10Gbps interface, replies can be sent from the 100Mbps e0M interface.

Similarly, when the e0M interface is in the same subnet as the default gateway, it is very likely that the default route will be associated with the e0M interface, causing outbound routed traffic (even if a reply to incoming traffic from a remote subnet) to use the 100Mbps e0M interface.

But what’s the solution?

Perform the following steps to resolve the issue:

  1. Take down the management interface and allow the default gateway or static routes to get mapped to another interface.
  2. Assign an IP address to the e0M management interface from a different subnet, one dedicated for management traffic and isolated from production traffic.

If the environment is such that the e0M interface cannot be re-assigned to a dedicated management subnet, then e0M should be unplugged and left unconfigured. In this case, management traffic should be allowed to run over the production network interfaces.



View all posts by