Moving a VIPA (for TCP/IP outage)

While a VIPA provides non-disruptive rerouting of IP data during failure of a physical interface, termination of the stack or the associated z/OS® (including planned outages) will disrupt connections or UDP sessions to applications on the terminated stack. While failure of the TCP connection or UDP session will be visible to the clients, the duration of the outage is determined by how long the client application is unable to reconnect to an appropriate server application. Because it is common in large enterprises to have multiple instances of an application on different z/OS images, if the VIPA address can be moved to another stack that supports the application, the clients can reconnect and the perceived outage will be over.

An IP address associated with a particular physical device is unavailable until the owning stack is restarted; however, a VIPA is not associated with any particular physical interface. If termination of a stack is detected and a suitable application already is active on another stack, the VIPA can be moved. Connections on the terminated stack will be disrupted, but they can be reestablished on the backup stack using the original VIPA.

Movement of a static VIPA to a backup stack can be accomplished by using VARY TCPIP,,OBEYFILE commands on the backup. The data set specified on the command must contain an appropriate set of DEVICE, LINK, HOME, and optionally, BSDROUTINGPARMS statements for IPv4 static VIPAs or INTERFACE statements for IPv6 static VIPAs. If OMPROUTE is used as the routing daemon, an appropriate interface statement is needed in the OMPROUTE configuration file. If the TCP/IP configuration file with the statements defining the VIPA is created in advance, the transfer can be accomplished via automation. This procedure is documented in z/OS Communications Server: IP Configuration Reference. Movement of a DVIPA, on the other hand, can be accomplished by configuring a stack to back up a specific DVIPA that is defined on another stack. In this case, failure of the defining stack causes the DVIPA to move without operator intervention or extra automation. See Planning for dynamic VIPA takeover for more information. Regardless of the type of VIPA to be moved, it is up to the system programmer or operator to ensure that the VIPA is moved to a backup stack that has the appropriate server applications.

In the absence of a failure, a VIPA is just like any other IP address, and routing for a VIPA is the same as for an IP address associated with a physical link.