Network requirements for high availability

The high availability configuration must meet minimum network requirements to operate effectively.

It is recommended that you locate the appliances in a high availability pair in the same data center (preferably in adjacent racks).

If you choose to locate the appliance in different data centers, you must be aware of the following limitations:
  • Performance degrades rapidly with increasing latency between data centers. Although IBM will support a latency of up to 10 ms (anything longer is not supported), you might find that your application performance cannot tolerate more than 1 to 2 ms of latency.
  • You must configure the primary and secondary Ethernet interfaces used for HA across completely redundant links (that is, do not rely on shared networking hardware, cabling, or power supplies for these connections). It is recommended that the replication interface is connected over a third redundant link.
  • The links must have sufficient dedicated bandwidth with no contention.
  • If an HA queue manager has an encrypted file system then the data is replicated in its encrypted form, otherwise the data sent across the replication link is not subject to any additional encryption beyond that which might be in place from using MQ AMS.
  • Be aware that if you lose the network connections between the two appliances, a partitioned situation can arise where the same queue manager continues to run on each appliance and each instance has a different set of queue manager data. When the connection is restored you must take action to specify which set of data you want to preserve, and which you want to discard. (This is sometimes called a 'split-brain' situation).

The primary, secondary, and replication links that are used for HA can either be connected to a switch, or they can be directly cabled between the appliances. If link aggregation is used for the HA replication link with directly cabled connections, then the link aggregation cannot be configured to use a type of active-backup.

The design of the network topology should be performed by networking experts with a deep understanding of the network architecture being employed. The tools ping and traceroute can be used as a quick way to begin to explore the network properties, but are not a substitute for a detailed review of the network architecture.

Use ping to test that you can connect to the other appliance in a high availability pair:
  1. Connect to the IBM® MQ Appliance as described in Command line access. Log in as an administrative user.
  2. Type the following command to test your connection: 
    
ping remote_IP_address
    Where remote_IP_address is the IP address of the eth13 interface of the other appliance in the high availability pair.
  3. Repeat the ping test with the IP addresses of the eth17 and eth21 interfaces of the other appliance in the high availability pair.
The ping command sends 6 Internet Control Message Protocol (ICMP) echo-request messages to the specified host with a one second interval between each message and reports the results.
Use traceroute to test the connection, reporting the addresses of any hosts used to make the connection, and the latency of the connection:
  1. Connect to the IBM MQ Appliance as described in Command line access. Log in as an administrative user.
  2. Type the following command to test your connection: 
    
traceroute remote_IP_address
    Where remote_IP_address the IP address of the eth13 interface of the other appliance in the high availability pair.
  3. Repeat the traceroute command with the IP addresses of the eth17 and eth21 interfaces of the other appliance in the high availability pair.
The traceroute command traces the route that packets actually take to their target host. The output shows the IP address of the hops (for example, gateway or routers) and the round trip time.

Because of the load it imposes on the network, do not use this command too often during typical operations.