TCP/IP (Parallel Library) Migration Guide

The New Socket Provider
HP NonStop TCP/IP (Parallel Library) Migration Guide—522272-003
2-4
Fault-Tolerant Behavior
Table 2-1. Fault-Tolerant Behavior; Conventional TCP/IP Process Compared to
TCPSAM
Failure Description
Conventional TCP/IP as
Transport-Service Provider
TCPSAM as Transport-
Service Provider
TCP/IP primary process
goes down
•
Backup process takes
over.
•
All TCP sockets in the
established state return
ECONNRESET to the
application and send
RST to the remote end.
•
Listen and UDP sockets
survive.
•
Backup takes over.
•
All sockets survive.
TCP/IP primary processor
goes down
•
Backup takes over.
•
All TCP sockets in the
established state return
ECONNRESET to the
application and send
RST to the remote end.
•
Listen and UDP sockets
survive.
•
Backup takes over.
•
All sockets survive if the
application is not on the
failed processor
because the sockets are
on the application
processor, and might not
be on the TCPSAM
processor.
TCP/IP backup process goes
down
•
All sockets survive.
•
Socket is
non-fault-tolerant until
the backup TCP/IP is
brought up.
•
All sockets survive.
TCP/IP backup processor
goes down
•
All sockets survive.
•
Socket is
non-fault-tolerant until
the backup TCP/IP is
brought up.
•
A non-fault-tolerant
application on that
backup processor goes
away.
•
All sockets survive (if the
application is not on the
failed processor).
Application primary dies
•
Backup application can
take over and continue
on the existing sockets.
•
Backup application,
upon takeover, receives
an error ECONNRESET
and has to close the
socket.