This information covers a problem that can occur with connectivity.
The problem and the solution are provided to help you complete problem
determination activities.
When the production server and backup server are separated
by a firewall, socket connections might time out
- Question
- After a successful snapshot backup operation, why is it not possible
to mount or unmount this snapshot backup on a backup or clone server?
- Answer
- The socket connection failure can result from a mismatch between
the firewalls connection timeout setting and the operating systems
frequency of sending keep alive network packets. When a firewall or
other network devices such as a router or switch exists between the
production and backup server, the daemon connection can time out.
A similar situation can exist between a production and clone server.
To prevent connections from timing out, the management agent acsd on
the production server, requests that the operating system sends out
network packets. These packets keep the connection between the servers
alive.
The
tcp_keepidle operating system parameter
specifies the interval of inactivity. Depending on the operating system,
this parameter might vary. After this interval of inactivity, the
TCP generates a keep alive transmission for the application that requests
it. This interval is measured in half seconds. For AIX® operating systems, the keep alive default
value for this parameter is 14400 (2 hours). This frequency is sufficient
for many environments. Decrease this value when the following conditions
exist:
- A firewall or other network device exists between the production
and backup or clone server.
- If the device connection timeout is less than 2 hours.
For AIX operating
systems, issue the following network command to reduce the
tcp_keepidle parameter
value and send a keep alive transmission every 5 minutes:
no -o tcp_keepidle=600
This
change remains in effect until you restart the production server.
To permanently modify this parameter, add the command to the
/etc/rc.net file.