Configuring Probe-Based Failure Detection

Probe-based failure detection involves the use of target systems, as explained in Probe-Based Failure Detection. In identifying targets for probe-based failure detection, the in.mpathd daemon operates in two modes: router target mode or multicast target mode. In router target mode, the daemon probes targets that are defined in the routing table. If no targets are defined, then the daemon operates in multicast target mode, where multicast packets are sent out to probe neighbor hosts on the LAN.

Preferably, you should set up target systems for the in.mpathd daemon to probe. For some IPMP groups, the default router is sufficient as a target. However, for some IPMP groups, you might want to configure specific targets for probe-based failure detection. To specify the targets, set up host routes in the routing table as probe targets. Any host routes that are configured in the routing table are listed before the default router. IPMP uses the explicitly defined host routes for target selection. Thus, you should set up host routes to configure specific probe targets rather than use the default router.

To set up host routes in the routing table, you use the route command. You can use the -p option with this command to add persistent routes. For example, route -p add adds a route that will remain in the routing table even after you reboot the system. The -p option thus enables you to add persistent routes without needing any special scripts to re-create these routes with every system startup. To optimally use probe-based failure detection, make sure that you set up multiple targets to receive probes.

The procedure How to Manually Specify Target Systems for Probe-Based Failure Detection shows the exact syntax to add persistent routes to targets for probe-based failure detection. For more information about the options for the route command, refer to the route(1M) man page.

The following topics are covered in this section:

Requirements for Choosing Targets for Probe-based Failure Detection

Follow this list of criteria when evaluating which hosts on your network might serve as good targets:

• Make sure that the prospective targets are available and running. Make a list of their IP addresses.

• Make sure that the target interfaces are on the same network as the IPMP group that you are configuring.

• The netmask and broadcast addresses of the target systems must be the same as the addresses in the IPMP group.

• The target host must be able to answer ICMP requests from the interface that is using probe-based failure detection.

Configuring Probe-Based Failure Detection (Task Map)

The following task map lists the procedures to configure probe-based failure detection for an IPMP group.

How to Select Which Failure Detection Method to Use

By default, probe-based failure detection operates by using test addresses. If the NIC driver supports it, link-based failure detection is also enabled automatically.

You cannot disable link-based failure detection if this method is supported by the NIC driver. However, you can select which type of probe-based failure detection to implement.

Before You Begin

Make sure that your probe targets meet the requirements listed in Requirements for Choosing Targets for Probe-based Failure Detection.

1. To use only transitive probing, perform the following steps:
   1. Use the appropriate SMF commands to switch on the IPMP property transitive-probing.

      # svccfg -s svc:/network/ipmp setprop config/transitive-probing=true
      # svcadm refresh  svc:/network/ipmp:default

      For more information about setting this property, see the in.mpathd(1M) man page.

   2. Remove any existing test addresses that have been configured for the IPMP group.

      # ipadm delete-addr address addrobj

      where addrobj must refer to an underlying interface that hosts a test address.

2. To use test addresses to probe for failure, perform the following steps:
   1. If necessary, turn off transitive probing.

      # svccfg -s svc:/network/ipmp setprop config/transitive-probing=false
      # svcadm refresh  svc:/network/ipmp:default

   2. Assign test addresses to the underlying interfaces of the IPMP group.

      # ipadm create-addr -a address under-interface

      where address can be in CIDR notation and under-interface is an underlying interface of the IPMP group.

How to Manually Specify Target Systems for Probe-Based Failure Detection

This procedure describes how to add specific targets if you are using test addresses to implement probe-based failure detection.

Before You Begin

Make sure that your probe targets meet the requirements listed in Requirements for Choosing Targets for Probe-based Failure Detection.

1. Log in with your user account to the system on which you are configuring probe-based failure detection.
2. Add a route to a particular host to be used as a target in probe-based failure detection.

   $ route -p add -host destination-IP gateway-IP -static

   where destination-IP and gateway-IP are IPv4 addresses of the host to be used as a target. For example, you would type the following to specify the target system 192.168.10.137, which is on the same subnet as the interfaces in IPMP group ipmp0:

   $ route -p add -host 192.168.10.137 192.168.10.137 -static

   This new route will be automatically configured every time the system is restarted. If you only want to define a temporary route to a target system for probe-based failure detection, then do not use the -p option.

3. Add routes to additional hosts on the network to be used as target systems.

How to Configure the Behavior of the IPMP Daemon

Use the IPMP configuration file /etc/default/mpathd to configure the following system-wide parameters for IPMP groups.

• FAILURE_DETECTION_TIME

• FAILBACK

• TRACK_INTERFACES_ONLY_WITH_GROUPS

1. Become an administrator.

   For more information, see How to Use Your Assigned Administrative Rights in Oracle Solaris 11.1 Administration: Security Services.

2. Edit the /etc/default/mpathd file.

   Change the default value of one or more of the three parameters.

   1. Type the new value for the FAILURE_DETECTION_TIME parameter.

      FAILURE_DETECTION_TIME=n

      where n is the amount of time in seconds for ICMP probes to detect whether an interface failure has occurred. The default is 10 seconds.

   2. Type the new value for the FAILBACK parameter.

      FAILBACK=[yes | no]

      • yes– The yes value is the default for the failback behavior of IPMP. When the repair of a failed interface is detected, network access fails back to the repaired interface, as described in Detecting Physical Interface Repairs.

      • no – The no value indicates that data traffic does not return to a repaired interface. When a failed interfaces is detected as repaired, the INACTIVE flag is set for that interface. This flag indicates that the interface is currently not to be used for data traffic. The interface can still be used for probe traffic.

        For example, assume that the IPMP group ipmp0 consists of two interfaces, net0 and net1. In the /etc/default/mpathd file, the FAILBACK=no parameter is set. If net0 fails, then it is flagged as FAILED and becomes unusable. After repair, the interface is flagged as INACTIVE and remains unusable because of the FAILBACK=no value.

        If net1 fails and only net0 is in the INACTIVE state, then the INACTIVE flag for net0 is cleared and the interface becomes usable. If the IPMP group has other interfaces that are also in the INACTIVE state, then any one of these INACTIVE interfaces, and not necessarily net0, can be cleared and become usable when net1 fails.

   3. Type the new value for the TRACK_INTERFACES_ONLY_WITH_GROUPS parameter.

      TRACK_INTERFACES_ONLY_WITH_GROUPS=[yes | no]

      ---

      Note - For information about this parameter and the anonymous group feature, see Failure Detection and the Anonymous Group Feature.

      ---

      • yes– The yes value is the default for the behavior of IPMP. This value causes IPMP to ignore network interfaces that are not configured into an IPMP group.

      • no – The no value sets failure and repair detection for all network interfaces, regardless of whether they are configured into an IPMP group. However, when a failure or repair is detected on an interface that is not configured into an IPMP group, no action is triggered in IPMP to maintain the networking functions of that interface. Therefore, the no value is only useful for reporting failures and does not directly improve network availability.

3. Restart the in.mpathd daemon.

   # pkill -HUP in.mpathd

   The daemon restarts with the new parameter values in effect.