|
|
|
This section presents several advanced cluster scenarios and procedures for their configuration.
The recommended cluster architecture contains interfaces connected to a Layer-2 segment that is isolated from other clusters.
However, in a deployment where multiple clusters need to connect to the same Layer-2 segment, the same MAC address may be used by more than one cluster for Cluster Control Protocol (CCP) communication. This may direct traffic to the incorrect cluster.
Follow sk25977.
Cluster members use CCP to communicate with each other. In order to distinguish CCP packets from ordinary network traffic, CCP packets are given a unique source MAC address.
Default Value Of Fifth Byte |
Purpose |
|---|---|
|
CCP traffic |
|
Forwarding layer traffic |
When multiple clusters are connected to the same Layer-2 segment, setting a unique value to the fifth byte of the MAC source address of each cluster allows them to coexist on the same Layer-2 segment.
To change a cluster's MAC source address, run these commands on each cluster member:
fw ctl set int fwha_mac_magic <value>
fw ctl set int fwha_mac_forward_magic <value>
Parameter |
Default value |
|---|---|
|
|
|
|
Use any value, as long as the two gateway configuration parameters are different. To avoid confusion, do not use the value 0x00.
You can configure the above parameters to persist following reboot.
fwkern.conf, located at $FWDIR/boot/modules/.Parameter=<value in hex>. Make sure there are no spaces.By default, ClusterXL only monitors two VLANS for failure detection and failover. These are the highest and lowest VLAN tags defined for a given interface.
For example, if the topology for interface eth1 includes several VLAN tags in the range of eth1.10 to eth1.50, ClusterXL only monitors VLANs eth1.10 and eth1.50 for failure. Failures on any of the other VLANs are not detected in the default configuration.
|
Note - The command |
When both the highest and lowest VLANs fail, all the VLANs are considered down, and a failover occurs. This means that if a VLAN which is not listed as the highest or lowest goes down, the trunk is still considered "up", and no failover occurs.
There are instances in which it would be advantageous to monitor all the VLANs in the trunk, not just the highest and lowest, and initiate a failover when any one of the VLANs goes down.
To enable monitoring of all VLANs, enable the fwha_monitor_all_vlan property in $FWDIR/boot/modules/fwkern.conf. Change the property to fwha_monitor_all_vlan=1.
The default ClusterXL Control Protocol transport mode is auto. Use this command to configure the broadcast or multicast mode for the cluster.
To configure the ClusterXL Control Protocol to broadcast or multicast transport mode:
Important - The ClusterXL Control Protocol transport mode must be the same on all Cluster Members. When changing the CCP transport mode on Cluster Members, they can temporarily stop detecting the CCP packets from each other. This can cause a cluster fail-over. Therefore, we recommend you schedule a maintenance window for this change.
show cluster interfaces virtual
cphaprob [-vs all] -a if
set cluster ccp {multicast | broadcast}
cphaconf set_ccp {multicast | broadcast}
show cluster interfaces virtual
cphaprob [-vs all] -a if
In VSLS VSX Clusters of version R80.10 and higher, changing the CoreXL configuration or running the cphastop command in the context of VS0 results in failover of all the Active Virtual Systems on that VSX Cluster Member. This behavior is by design. Active Virtual Systems on other VSX Cluster Members do not fail over.
