unwanted tcp persistance

[fishiface]

I'm running a database server on an AIX 4.3 box. The database system spawns one process per connection. Some of these processes are running long after the client has given up and gone home. Here's one that is heading towards it's first birthday:

[b]ps -fp 11584[/b]
  UID    PID   PPID  C    STIME TTY  TIME CMD
 rsmi  11584  79582  0   Nov 18   -  0:00 /usr/informix/lib/sqlexec stx_se po

I'm not blaming the database server because it is being told that the connection is still active:

[b]lsof -p 11584 -a -i TCP[/b]
COMMAND   PID USER FD TYPE     DEVICE SIZE/OFF NODE NAME
sqlexec 11584 rsmi 7u IPv4 0x702f86dc      0t0  TCP po:sql->rsmi-vpn:1066 ([blue]ESTABLISHED[/blue])

I've checked that tcp has sensible settings:

[b]no -a | grep tcp[/b]
        tcp_ndebug = 100
           tcp_ttl = 60
     tcp_sendspace = 16384
     tcp_recvspace = 16384
     [blue]tcp_keepintvl = 150[/blue]
      [blue]tcp_keepidle = 14400[/blue]
       tcp_mssdflt = 512
      [blue]tcp_keepinit = 150[/blue]
 tcp_pmtu_discover = 1
      tcp_timewait = 1
 tcp_ephemeral_low = 32768
tcp_ephemeral_high = 65535
       tcp_newreno = 0
   tcp_nagle_limit = 65535

At which point I really don't know where to go next. I can (obviously) kill the processes but would rather leave them for debugging until I get a solution to stop the problem occurring in the first place. The clients are a mixture of Linux servers, LAN- and WAN-attached windows desktops and even, as in this example, a modem-attached VPNed laptop so I don't think it's a client problem.

Has anyone come across this problem or got any ideas as to how to proceed?

Yours,

fish

["]As soon as we started programming, we found to our surprise that it wasn't as easy to get programs right as we had thought. Debugging had to be discovered. I can remember the exact instant when I realized that a large part of my life from then on was going to be spent in finding mistakes in my own programs.["]
--Maur

 

[alexhu]

I have a similar problem on AIX 4.3.3 with Oracle.

However, sometimes these processes cause me grief, by going into a spin and consuming lots of cpu.

So I just kill them.

Running a 6 year old app, on out of date Oracle (8.0.5) and desupported AIX. I don't think I'm going to find a solution - most vendors will just say upgrade to a modern version of the software (not possible in this case)

Alex

 

[fishiface]

Thanks for the sanity-check. I'm in a similar situation using Informix on the AIX4.3 box: both are being replaced with Oracle on Linux within a year so there's no budget (or real interest) in upgrading either. I'm glad that I'm not missing anything obvious, though.

Yours,

fish

["]As soon as we started programming, we found to our surprise that it wasn't as easy to get programs right as we had thought. Debugging had to be discovered. I can remember the exact instant when I realized that a large part of my life from then on was going to be spent in finding mistakes in my own programs.["]
--Maur