I did notice other LynxOS processes such as NFS and other services
running. These seemed to have priorities around the 15-20 range. I'm
wondering if they might have been what was interferring with the low
priority thread in our application.
Or application is running fine now, but I would still like to
determine the cause of the original problem. Any comments or
suggestions appreciated.
So another thread in the system somewhere is consuming all available
CPU time. Remember that LynxOS is a strict priority system; if a
thread at priority 20 is always ready, no thread below that will get any
cycles at all[*].
> Note that even at
>40 it was still the lowest priority thread in our application. If
>it's of any interest, this thread does use a tcp socket.
It matters a little. What priority is hbtcpip (assuming it's still
called that) running at? Is it a disk-based system? Is swapping
enabled? Did your app perhaps get paged out?
>I did notice other LynxOS processes such as NFS and other services
>running. These seemed to have priorities around the 15-20 range. I'm
>wondering if they might have been what was interferring with the low
>priority thread in our application.
Foggy memories tell me they were probably running at 17.
Yes, that's almost certainly what is going on. Some strategic
ps checks should tell you what thread's getting all the time.
>Or application is running fine now, but I would still like to
>determine the cause of the original problem. Any comments or
>suggestions appreciated.
Figure out what process was getting all the time, and then figure
out why. Obvious examples include an NFS server getting hammered
by NFS clients, a web server in the same position, etc.
If your thread absolutely needs to run, make sure its priority
is such that it will get the time.
[*] OK, priority ceiling and/or priority inheritance aside.
--
Steve Watt KD6GGD PP-ASEL-IA ICBM: 121W 56' 57.5" / 37N 20' 15.3"
Internet: steve @ Watt.COM Whois: SW32-ARIN
Free time? There's no such thing. It just comes in varying prices...