Re: unable to create virtual circuit b/c "Too many open files"

[Andrew Johnson <anj@aps.anl.gov>]

On Sunday 06 July 2008 12:06:10 Bertrand H.J. Biritz wrote:
> I should then implement a portion which checks the connection status
> to the PV's, correct?

Well you already pass your connectionChangedCB() function to the 
ca_create_channel() routine and that callback should be checking to make sure 
that it's being told about a connection coming up rather than going down.  
You are calling ca_clear_channel() which should be tidying up everything 
related to the channel, but then you just loop back around to make the 
connections again, which rather pointless.

You will reduce the load on all your IOCs and your client machine if you make 
the connections once and keep them up the whole time, as long as you're using 
ca_get() or ca_get_callback() rather than ca_create_subscription() to fetch 
the data â if you subscribe you'll get all the data from those PVs, which may 
load your IOCs unnecessarily if the PVs process at any significant rate.

There are two fundamentally different ways to structure this kind of code, 
using callbacks or procedurally, and I suspect your best option in this case 
would be the procedural approach, which is probably easier to follow.  That 
would use a structure like this (adding status checks etc.):

ca_context_create();
for (...) {
    ca_create_channel(pv[i], NULL, ...);
}
ca_pend_io(10.0);
while(!done) {
    int getting[12];
    for (...) {
        if (ca_state(pCh[i]) == cs_conn) {
            ca_get(DBR_..., pCh[i], &value[i]);
            getting[i] = 1;
        } else {
            getting[i] = 0;
        }
    }
    ca_pend_io(5.0);
    for (...) {
        if (getting[i]) {
            // write value[i] to file
        } else {
            // write pv[i] not connected to file, if desired.
        }
    }
    ca_pend_event(5.0*60); // 5 minute delay
}
for (...) {
    ca_clear_channel(pCh[i]);
}
ca_context_destroy();

- Andrew
-- 
Talk is cheap. Show me the code. -- Linus Torvalds