We've two nodes running on grid for our Informatica enviornment and recently upgraded to 9.1 HF3.Since after that we're having sessions terminated which are running on non-master gateway node.Below are the error logs from Admin console.The domain process is making the node as disabled even if its up and running due to its not able to send update with in interval time.But the node is nt fully utilized at that time.Even assuming there is network latency and IO issues while writing to shared directory on the servers.How we can make the domain stable in this scenarios.The network team informed that there is no issues with the network.
2012-03-29 03:31:26 : INFO : (18336 | 32) : (IS | STG_INT_) : NODE_2 : LB_47063 : Detaching DISABLED node [NODE_2] with  started requests and  reserved requests.
2012-03-29 03:31:26 : INFO : (18336 | 32) : (IS | STG_INT) : NODE_2 : LM_36956 : Oubound connection to node [NODE_2] lost
How to find the issue in this senarios.
Maybe you've run out of tcp ports?
On win 2003 server and below you need to configure max port number range to 65000, or even a few sessions per second (say 1) will make the system highly unstable.
Have the upgrade been done 'in place' on the same servers/os as before or was anything else changed during the upgrade process?
We're using Sun Solaris box for Informatica host.How to check if the tcp ports ran out of range?
We're upgraded the same servers from 8.6 to 9.1.
Anil Kumar Kathraji
I'd talk to my solaris sys-adm, and ask him to measure it...
This is what Google came up with after 2 minutes of searching: