Availability Guide for Application Design

Increasing the Availability of Tuxedo Applications
Availability Guide for Application Design—525637-004
5-10
Availability of NonStop Tuxedo Applications
Whether the TCP/IP process has a backup process or not, all connections to
workstation clients through the failing TCP/IP process are lost. In addition, any
currently active TMF transactions initiated by the WSH process on behalf of
workstation clients are aborted. Workstation clients must detect errors, reconnect, and
restart the failed transactions.
Recovery From BBL Process Failure
The application is unaffected by the loss of a BBL process except for the function
provided by the processor in which the BBL runs.
The BBL is not restarted in another processor, because each BBL process is specific
to a given processor; if the processor fails then its BBL has no reason to exist.
When the processor is reloaded, or if the BBL unexpectedly stops for a reason other
than processor failure, the BBL process is restarted by the supervisor process on
request from the DBBL process. The BBL creates its internal state information from the
local shared memory file (which it previously created) or from information provided by
the DBBL.
Recovery From DBBL Process Failure
The DBBL process runs as a NonStop TS/MP server class. If the DBBL stops, the rest
of the application is unaffected, except that functions involving the DBBL, such as
starting a server process, cannot be accessed.
The DBBL is restarted immediately by the supervisor process in a processor selected
by the supervisor process. The new DBBL creates its internal state information from
the BBL in the local processor.
Recovery From Supervisor Process Failure
The supervisor process always runs as a NonStop process pair. If the primary process
fails, the backup takes over in the processor in which it is configured to run. Because
the supervisor process checkpoints all the information it needs to continue managing
the application, there is no effect on the application.
When a failed processor is reloaded, you can use the xtuxadm graphical interface
utility to return to the original process configuration.
Recovery From Client Process Failure
If a native client process fails, any TMF transactions initiated by the client process are
aborted. The client process must be manually restarted.
If a workstation client process fails, the consequences depend on whether the client
process uses dynamically or statically bound /WS libraries. If the client uses
statically-bound /WS libraries, then the WSH process detects an error as the TCP/IP
connection is unexpectedly closed; the WSH process aborts any outstanding TMF
transactions started on behalf of the client process.