Running OnTap 9.3. Is there a way to increase the TCP window size for snapmirror connections?
I used ::*>network connections options buffer modify -receive-buffer 7168 -network WAN -protocol tcp -service ctlopcp and increased the buffer size. Understand there are 12 tcp connections per node. When I run multiple snapmirrors at once I get over 200MB/s. When I run a single mirror from source node to destination node, I only see about 35MB/s.
Could you please check if there is any throttle applied on the snapmirror relation level. ::snapmirror show -destination-path -fields throttle if yes please set it to "unlimited" and retry the snapmirror update.
Also could you let us know how is above throughput number are determined? Snapmirror performance troubleshoot would require deep investigation including capture of packet trace and perf stats logs. to troubleshoot further i would request you to open a support ticket with us so we can look into it.
Very late reply. But my snapmirror throughput dramatically increased to utilizing the maximum amount of bandwidth on the WAN connection after creating new intercluster LIFS on a all-new VLAN, then creating a separate IP-Space and broadcast domains with those vlan ports in the broadcast domains. Isolating the intercluster traffic into their own IP spaces.
Per support.. See these articles:
How to create and configure Cluster Peering with designated Intercluster LIFs using IPSpaces