User guide

Appendix C - Redundancy C-5
VMS Redundancy
Automatic Synchronization
As the name implies, automatic synchronization occurs automatically by the
active VMS and is performed whenever any changes occur that are associated
with automatic system functions, such as automatic switching, device redun-
dancy, etc. The active server maintains a memory cache that holds the updates
until they can be pushed out to the standby servers by an automatic synchroniza-
tion that occurs during the VMS heartbeat. The updates are tagged onto the
heartbeat message that is sent by the active server to the standby servers.
Manual Synchronization
Manual synchronization, also referred to as “full synchronization”, must be
performed by administrator/user command for any changes not related to auto-
matic VMS functions, such as whenever any database configuration changes are
made to the server. Should a standby server be restarted, when it rejoins the
redundancy group, the sequence of updates is lost and a manual synchronization
is required to ensure that the standby receives the most current database from
the active server.
During a full synchronization, the active VMS service is temporarily taken
down to avoid any changes occurring during the synchronization process. The
active server sends the contents of the temp file holding the entire database
backup to each standby server via simultaneous unicasts. If, for any reason,
there is a failure with this update process, a notification will appear in the
windows log.
Server Contention
Server contention is a built-in protection mechanism for redundant VMS opera-
tion. A situation may occur where the active server briefly loses network
connectivity—a network cable is unintentionally pulled, for example—before
communications are restored. The first priority standby will become active due
to the lost heartbeat of the former active server. When the former active server
returns, it will detect that there is another active server in operation, and will
enter the contention state.
When this is sensed by the current active server, it also will enter the contention
state. In such a situation, there is no way for the system to determine which
server has the most current up-to-date database, and both servers will immedi-
ately de-activate to protect the current status of the network. A generated alarm,
both visual and audible activated, will appear on each server. In addition, an
SNMP trap will be generated.
In this condition, VMS services are still running, but no changes of state can be
executed in the network until the condition is cleared. For instructions on clear-