Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Wed, 16 Nov 2005 15:07:25 -0600

Marty Kraimer wrote:

So why did we invent epicsAtExit?
Does anyone remember?

I think it was to try and ensure that vxWorks IOCs shut down "nicely"

when rebooted with a Ctrl-X. The CA server gets a chance to close all

its sockets, thus the clients are immediately notified that the IOC is

going down. Without it, the clients don't find out until up to 35

seeconds later that the IOC is no longer there. Where an OS such as

Linux carries on running after a soft IOC shuts down it immediately

closes all of the CA server's TCP sockets as part of the process

tidy-up, so this problem doesn't appear on soft IOCs.

If a vxWorks IOC reboots up before a client even notices that it ever

went down, there could also be extra complications before it reconnects,

since the client's network stack just thinks that the TCP socket was

congested for a while but is still live; the network stack in the

restarted server will receive packets intended for the old socket and

will have to reject them. I don't know if there are any additional

timeouts that this might introduce.

I think there may also have been issues with shutdown on WIN32 systems,

the atExit() routines were being run after some DLLs had been unloaded

or something like that. I may be totally wrong about this though.

- Andrew
--
* * Matt Santos / / Leo McGarry * * For a Brighter America * *

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System