How to handle CLOSE_WAIT

[byk]

Hi,

We have a client-server based application comunicating over sockets. Dropping the connection from client side is forcing the server-child-process to a CLOSE_WAIT (netstat) state.
Please suggest how to get rid of this and close the server-child-process.

Regards,
byk

 

[JasonDeckard]

Normally, a TCP session is concluded with a brief message exchange (much like the 3-way handshake):

System A       System B
--------       --------
 FIN     ---->
         <----  FIN/ACK
 ACK     ---->

CLOSE_WAIT occurs when the remote system does not finish the hang-up process:

System A       System B
--------       --------
 FIN     ---->
         <----  FIN/ACK
        (NO ACK)

In this case, system B will resend the FIN/ACK a few times before timing out. The state of the socket from System B's perspective is CLOSE_WAIT until System A either acknowledges the FIN/ACK, or System B times out (gives up).

CLOSE_WAIT is a normal state for a socket, but if everything goes smoothly this state is short-lived. CLOSE_WAIT is not cause for alarm, and can be safely ignored.

What can you do about CLOSE_WAIT when the remote system does not finish the hang-up? Not much, other than wait for the connection to time-out.


Best of luck,
Jason Deckard

 

[byk]

The server socket is not timing out even after 2 hours.

Is there any way of getting the pid of the process which holds the CLOSE-WAIT socket (either through ps or netstat etc) ? If so that the process-id can be killed by a monitor process.

Please advise.

Regards,
byk

 

[JasonDeckard]

I agree that two hours is excessive.

Although the operating system controls how long the socket will remain in CLOSE_WAIT, you can find out which process is controlling the port with lsof (the source is at ftp://vic.cc.purdue.edu/pub/tools/unix/lsof/):

lsof -i | grep (port number)

 

[byk]

It seems that "lsof" is not a standard HP-UX command.
Moreover in this case I have multiple child processes listening over the same socket port.
How to determine the pid of the process that has actually
reached CLOSE_WAIT

By the way .. thanks for the replies Jason.

Regards,
byk

 

[byk]

Hi,

Please suggest me the syntax for using a setsockopt in C over HP-UX. What exaxtly are the parameters 4,5 in setsockopt?
int setsockopt(int s,int level, int optname, const void* optval, int optlen);

I actually need to try SO_KEEPALIVE to solve the problem mentioned.

Regards,
byk

 

[JasonDeckard]

Parameter four is the value of the setting (for the option named in paramter three). Parameter five is the length, in bytes, of the value provided in parameter four:

  int trueval = 1;
  rval = setsockopt( sockid, SOL_SOCKET, SO_KEEPALIVE, &truval, sizeof(trueval) );

I'm curious to find out if setting the keep-alive solves your problem. I would be suprised if your OS attempts to send a keep-alive packet from a socket that is not in the CONNECTED state.

Good luck,
Jason

 

[iambosley]

We have used the command :

ndd -set /dev/tcp tcp_discon_by_addr
"0a0a04081f42 a0a8c2f0527"

Where :

local ip=10.10.4.8 = 0x0a0a0408
local port=8002 = 0x1f42

remote ip = 10.10.140.47 = 0x0a0a8c2f
remote port = 1319 = 0x0527


Roger Bosley
Capital Blue Cross