Virtual IP for service instances

To prevent service instances from sharing IP addresses, you can configure a virtual IP (VIP) address for a service, so that each service instance is assigned a separate VIP.

With VIP configured, each service instance has a separate VIP, and the number of the VIPs is determined by the maximal number of the service instances that the service could have. Since there is one base VIP, the VIPs for one service are continuous, which means that discontinued VIP settings are not supported. All service instances must be started within the same subnet, so that the host physical IPs (at least in one network interface) and VIPs are in the same subnet.

To support VIP for a service, the VIP address is applied to the target host where the service instance is going to run. If you start one service that has a VIP address configured, each running service instance has one specified VIP address.

Important: To configure a virtual IP for a service, ensure that the following conditions are met:
  • All the hosts on which the service might run must be in the same subnet (at least one network interface). The virtual IPs must also be in the same subnet.
  • The virtual IPs must be reserved.
  • If a firewall is configured inside the subnet, the IP addresses must be accessible. To prevent IP conflicts, configure the ServiceFailoverFencing parameter. See Service profile reference.

To configure virtual IPs for a service from the cluster management console, you must be a cluster administrator or have the Services Manage permission.

With virtual IP, each service instance has a separate virtual IP that is applied to the host on which the service instance is to run. If you start one service with a virtual IP address, each running service instance has one specified virtual IP address.

The number of virtual IPs is determined by the maximum number of service instances that the service can have. Because there is one base virtual IP, the virtual IPs for one service are continuous (in other words, discontinued virtual IP settings are not supported). All service instances must be started within the same subnet, so that the host physical IPs (at least in one network interface) and virtual IPs are in the same subnet.

A virtual IP is an IP address that is assigned to multiple applications that reside on a single server, multiple domain names, or multiple servers (rather than to a specific single server or network interface card). Incoming data packets are sent to the virtual address and routed to actual network interfaces.
Note: If you are setting up virtual IPs for the WEBGUI service, you must add the virtual IP host name into the parameter AllowlistRESTOrigin in the $EGO_CONFDIR/../../gui/conf/pmcconf/pmc_conf_ego.xml.

Limitations

Be aware of the following notes and limitations when configuring a virtual IP for a service in the CLI:
  • You must be a cluster administrator or have the Services Manage permission to configure virtual IPs in the CLI or in the service profile.
  • Users that have the All Clients View permission are able to retrieve the real host name by using the egosh client view command in the CLI.
  • Users that have the Spark Instance Groups View permission are able to retrieve the real host name in the cluster management console.
  • Users are able to retrieve the real host name by using the egosh resource list command in the CLI, or from the host list in the cluster management console.
  • Users are able to retrieve the real host name by using the egosh service list command in the CLI, or from the service list in the cluster management console.