[tipc-discussion] seeing a "broken pipe" error for intra-node TIPC connections

[Debnath G. <Deb...@bi...>]

Hi, 

 I am seeing "broken pipe"  error while doing a TIPC socket "send" on our system. 

 I have tried searching through different repositories on TIPC site just in case there were discussions related to this issue, but didn't find anything on the exact nature of the issue I am seeing. Apologies however, if there IS something already and I'd appreciate very much if you can point me to that.

 So now the details of the problem:

 I am using TIPC version 1.5.12 on Linux kernel version 2.6.14, and seeing the issue for TIPC connections  between applications running on same node ( i.e. intra-node TIPC connections ).

 And the issue is: sometimes when I do a socket "send" on an already established connection, the send fails and returns a "broken pipe" ( errno of EPIPE) error. [To my best of knowledge no other application thread or process is closing down the connection during the send ]

 Now, the issue is not very frequent and - my suspicion is - happens when the cpu resource is stretched for availability leading to cpu starvation for some of the system and user application processes, though I don't have data to prove that.

 Going through the materials on troubleshooting and FAQ on TIPC site, initially it seemed that the "Link Instability" could be what is going on here, but I was not sure if the "link" applies to intra-node TIPC connections too, which is what I have. Neither I am seeing any "peer not responding" messages when the issue happens.
 

 Questions: Any idea if this is a known issue ( for 1.5.12 version ) ?

 	    Otherwise, what if any, TIPC debugging facilities I can use 
	    to narrow down the problem to something more understood from TIPC
	    functionality point of view ?

Appreciate any comments, thoughts!

Thanks!