You can use hot standby distribution to have one preferred
target server and one or more backup (hot standby) target servers.
The preferred target server receives all new incoming connection requests
and the hot standby servers act as backups in case the preferred target
server becomes unavailable.
Before you begin
You need to determine which server should initially receive
the workload (the preferred server), and which server or servers should
be a backup server.
Procedure
Perform the following steps to configure hot standby
distribution on the VIPADISTRIBUTE statement:
- Configure HOTSTANDBY on the DISTMETHOD parameter.
- Configure the AUTOSWITCHBACK parameter or the NOAUTOSWITCHBACK
parameter.
- Configure AUTOSWITCHBACK if you want the distributor to automatically
switch distribution back to the preferred target when it is available.
For example, if the preferred target becomes a standby target because
its server is no longer ready, when the server is again in the listening
state, the distributor automatically switches back to the preferred
target as the active target. This is the default value.
- Configure NOAUTOSWITCHBACK if you want the distributor to continue
to use a backup target when the preferred target is available.
Rule: If NOAUTOSWITCHBACK is configured, then
the active server is initially determined by order of activation;
the first ready listener becomes the active server regardless of the
configured server type (PREFERRED or BACKUP), and remains the active
server unless it becomes unavailable.
- Configure the HEALTHSWITCH parameter or the NOHEALTHSWITCH
parameter.
- Configure HEALTHSWITCH if you want the distributor to automatically
switch from the active target when it is not healthy. This is the
default. A target is not healthy when there is a severe problem detected
by one of the following health metrics:
- Target server responsiveness (TSR) value is 0%
- Rate of abnormal transaction completions is 1000
- General health of the application is 0%
For more information, see Target server connection setup responsiveness monitoring, and the WLM health
metric information in SERVERWLM - Distribution using WLM server-specific weights.
Rule: When a switch occurs because a target
is no longer healthy, the target is not used as a backup target even
if its health recovers, unless all available backup targets have previously
had health problems. Health metric indicators typically recover when
a server is no longer receiving any work, and although a server can
appear healthy, the distributor cannot determine if the server is
actually healthy. When the distributor detects that the server is
transitioning back to the ready state, it again uses the target as
a backup.
Tip: You can use the VARY TCPIP,,SYSPLEX,QUIESCE
command and the VARY TCPIP,,SYSPLEX,RESUME command to manually bring
a server back to the ready state. You can also use these commands
before and after planned maintenance, or to temporarily divert new
workload requests from a particular target. For more information,
see
Manually quiescing DVIPA sysplex distributor server applications.
- Configure NOHEALTHSWITCH if you want the distributor to ignore
health metrics. The distributor switches from the active target only
when it is not ready or when the distributor does not have an active
route to the target.
- Configure the PREFERRED option or the BACKUP option after
each XCF address on the DESTIP parameter to designate the preferred
server and the backup servers, and configure a rank value for each
backup server.
- Configure PREFERRED to designate the preferred server. This must
be configured for one, and only one, XCF address.
- Configure BACKUP to designate a backup (hot standby ) server.
This must be configured for at least one XCF address. Also configure
a rank value on the BACKUP option for each backup server. The distributor
switches to the highest ranked backup if the preferred server becomes
unavailable.