HA terminology
HA cluster
Group of two devices instructed to work as a single entity. Every HA cluster has one
primary device and one auxiliary device. The primary device controls how the cluster
operates. The roles that the primary and auxiliary devices play in the cluster depend on the
configuration mode.
HA active-passive configuration mode
An HA cluster consists of a primary device and an auxiliary device. In this mode, only
the primary device processes traffic while the auxiliary device remains in stand-by mode,
ready to take over if a primary device failure occurs.
HA active-active configuration mode
An HA cluster consists of a primary device and an auxiliary device. In this mode, both
devices process traffic, and the primary device load-balances the traffic. The decision to
load-balance the traffic is made by the primary device. The auxiliary device can take over
this function if the primary fails.
Primary device
In an active-active cluster, the primary device receives all network traffic and acts
as load balancer to redirect traffic to the auxiliary device. The primary device also tracks
the status of all cluster devices. In an active-passive cluster, the primary device
processes the network traffic while the auxiliary device does not process any traffic but
remains ready to take over if the primary device fails.
Auxiliary device
In an active-active cluster, the auxiliary device processes the network traffic
assigned to it by the primary device. If the primary device fails, the auxiliary device
becomes the primary device. In an active-passive cluster, the auxiliary device does not
process network traffic and is in stand-by. It becomes active only when the primary device
is not available to process the traffic.
Dedicated HA link port
Dedicated HA link is a direct physical link between the devices participating in HA
cluster.
Load balancing
The ability of HA cluster of balancing the traffic between nodes in the HA
cluster.
Monitored interface
Set of interfaces that are selected to be monitored. Each device monitors its own
selected interface(s) and if any of them goes down, the device removes itself from the
cluster and a failover occurs.
Virtual MAC
It is a MAC address associated with the HA cluster. This address is sent in response
when any of the machines make an ARP request to HA cluster. It is not the actual MAC address
and is not assigned to any interface of any unit in the cluster.
The primary device owns
the MAC address and is used for routing network traffic. All external clients use this
address to communicate with the HA cluster. In case of failover, the new primary device
will have the same MAC address as the failed primary device. The cluster device which
has a virtual MAC address acts as a primary device.
Primary state
In active-active mode, the device that receives all traffic and performs
load-balancing is said to be in
primary state. A device can be in primary state
only when the other device is in auxiliary state.
In active-passive mode, the device in
charge of processing all the traffic is said to be in the primary state.
Auxiliary state
In active-active mode, the device that receives the traffic from the primary device is
said to be in
auxiliary state. A device can be in auxiliary state only when the
other device is in primary state.
In active-passive mode, the device that is not
processing the traffic is in auxiliary state. A device can be in auxiliary state only
when the other device is in primary state.
Standalone state
A device is said to be in standalone state when it is able to process traffic and when
the other device is unable to process traffic (for example, if it is in a fault state or
inoperative).
Fault state
A device is in fault state when it cannot process network traffic if a device or link
fails.
Peer
Once the HA cluster is configured, cluster devices are termed as peers i.e. for the
primary device, the auxiliary device is its peer device and vice versa.
Synchronization
The process of sharing the various cluster configuration, between cluster devices (HA
peers). Reports generated are not synchronized.
Link uptime (3 seconds)
Time taken by the dedicated link or monitored port to come up.
Heartbeat (keep-alive) interval (250 milliseconds)
Interval between heartbeat packet exchange by HA peers to confirm that the cluster is
functioning.
Device failover
If an device does not receive any communication within the predetermined period of time
from the HA peer, the peer device is considered to have failed. This process is termed
as device failover as when this occurs, the peer device is taken over.
Device failover detection time (peer timeout) (4 seconds)
When the primary device stops sending Heartbeat packets, it is declared dead at the
end of four seconds (250 milliseconds x 16 timeouts).
Note The peer is considered active if
a Heartbeat is received within 14 timeouts.
A failover is triggered seven
seconds after the cluster has come up (3-second link uptime + 4-second device failover
detection time). You can’t change the failover threshold.
Link failover
Both the device in an HA cluster continuously monitor the dedicated HA link and the
interfaces configured to be monitored. If any of them fails it is called link
failure.
Session failover
Whether it is a device or link failover, session failover occurs for forwarded TCP
traffic except for the virus scanned sessions that are in progress, VPN sessions, UDP, ICMP,
multicast, and broadcast sessions and proxy traffic.
Device normally maintains session
information for TCP traffic which is not passing through proxy service. Hence, in case
of failover, the device which takes over will take care of all the sessions (TCP session
not passing through proxy application). The entire process is transparent for the end
users.