Many problems encountered with Data Protector can be helped by using the KeepAlive functionality. These include Disk Agent and Media Agent timeouts, and backups seeming to ‘Complete’, but, never actually finishing. You may even see Session Statistics printed, but don’t get the popup when the session is completed
The keepalive parameter just turns on the system mechanism for keeping a network connection alive for all the connections we make
The keepalive packets are part of the TCP protocol on machines, and are transparent to the program opening or using the connection. They are handled by the system and are not seen by application. They normally default to 7200 seconds, or 2 hours.
From a HPUX Operating System perspective, check the link
http://www.filibeto.org/unix/hp-ux/lib/kernel/perftun/tcp_ip-performance-wp-c02020743.pdf pg 16
On the server where you are seeing the problem, edit the file
/opt/omni/.omnirc The ‘dot’ is part of the filename
This file does not exist by default. If you have this file, you can edit it, and, if not, you can create it using the HPUX ‘touch’ command
touch /opt/omni/.omnirc
Either way, add this line to the file:
OB2IPCKEEPALIVE=1 Activates the KeepAlive mechanism
After making the changes to the ‘omnirc’ file, save the file, making sure that it has no extension, like ‘txt’ or TMPL’
You can add this file on any server which is showing a problem, either a Disk Agent or a Media Agent. Generally speaking, the changes do not have to be made on the Cell Manager, unless it is also a Media Agent having this issue
So, for example, after adding the KeepAlive switch to the ‘omnirc’ file, you need to set the kernel parameters to change the default value of 7200 seconds.
From the command prompt on any UNIX server. “ndd” can be used to get/set the kernel parameter for keepalive packets interval
ndd -get /dev/tcp tcp_keepalive_interval
will report the interval in milliseconds. To change this, run the command
ndd –set /dev/tcp tcp_keepalive_interval [milliseconds]
So, to set this to 15 minutes,
ndd –set /dev/tcp tcp_keepalive_interval 900000
and run this again to be sure the parameter is set correctly
ndd -get /dev/tcp tcp_keepalive_interval
15 minutes is a reasonable value. The key is to set these Values low because most firewalls, etc. timeout after an hour or so. Having the interval set to 15 or even 30 min. Should not hurt anything.
You may have seen some reference to the ‘omnirc’ environment values
OB2IPCKEEPALIVETIME
OB2IPCKEEPALIVEINTERVAL
It needs to be emphasized that OB2IPCKEEPALIVETIME and OB2IPCKEEPALIVEINTERVAL are used on Windows servers only, and have no effect on HPUX or Linux servers
If you want to add this line to the .omnirc’ file proactively, and set the kernel parameters,, that is, without seeing the problem, this is acceptable to cut off a possible future problem
Up to this point, we have seen no negative effects to setting KeepAlive on HPUX servers