Solved: Cluster Peer Unavailable - Data and ICMP Reachable

nicholsongc · ‎2017-08-01

Attempting to peer two clusters (Cluster #1 and Cluster #2) but cannot get past Availability = Unavailable

Cluster #1 is a six (6)-node cluster

Cluster #2 is a two (2)-node cluster

Tried peering with and without authentication - current peering has Authentication - OK

Both Data and ICMP Ping status are session_reachable and interface_reachable respectively on both sides of the peering

All ports involved/recognized have the same MTU (9000)

Cluster #1 is currently successfully peered and the source to a 3rd-party provider for DR (SnapMirror)

Cluster #1 displays both nodes for Cluster #2 for Remote Cluster Nodes - output of cluster peer show -instance

Cluster #2 only shows Cluster #1's node1 (of six) in for Remote Cluster Nodes - output of cluster peer show -instance

Cluster #2 DOES, however, show all six IP addresses for each node of Cluster #1 for Active IP Addresses - output of cluster peer show -instance

nicholsongc · ‎2017-08-16

So the solution to my problem here was two-fold:

1. Existing Peer relationship between our production ClusterA and our DR provider's ClusterB was configured using ClusterA's DEFAULT IPSpace. On the advice of NetApp support I created a separate (dedicated) IPSpace for ClusterA to talk to ClusterC. This means I had to create a new/dedicated IPSpace as well and add a route.

2. MTU sizes. My downfall here was that my Intercluster LIFs on ClusterA and associated Broadcast domain is set to 9000. However, I confirmed with our DR provider that the ports/broadcast domain at their end is set to 1500. So rather than heed all advice I've ever read/seen here I assumed that if mis-matched MTUs worked for the ClusterA-ClusterB peering it should work for ClusterA-ClusterC. So I only changed ClusterC's broadcast domain and ports to run at MTU 1500. NO GO. When I changed my new broadcast domain on ClusterA and associated ports to 1500 the peering worked.

View solution in original post

nicholsongc · ‎2017-08-16

So the solution to my problem here was two-fold:

1. Existing Peer relationship between our production ClusterA and our DR provider's ClusterB was configured using ClusterA's DEFAULT IPSpace. On the advice of NetApp support I created a separate (dedicated) IPSpace for ClusterA to talk to ClusterC. This means I had to create a new/dedicated IPSpace as well and add a route.

2. MTU sizes. My downfall here was that my Intercluster LIFs on ClusterA and associated Broadcast domain is set to 9000. However, I confirmed with our DR provider that the ports/broadcast domain at their end is set to 1500. So rather than heed all advice I've ever read/seen here I assumed that if mis-matched MTUs worked for the ClusterA-ClusterB peering it should work for ClusterA-ClusterC. So I only changed ClusterC's broadcast domain and ports to run at MTU 1500. NO GO. When I changed my new broadcast domain on ClusterA and associated ports to 1500 the peering worked.