Brian Candler writes:
On Tue, Dec 21, 2004 at 04:57:37PM -0500, oth...@freeshell.org wrote:
This is just noting a behavior. These processes have been running since about
3PM Friday, long since the user logged out.
PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
5171 user 10 4 2024K 2016K wait 46:22 0.00% 0.00% sqwebmaild
5168 user 10 4 2024K 2016K wait 46:21 0.00% 0.00% sqwebmaild
5240 user -6 0 9008K 9712K pipewr 1:15 0.00% 0.00% ispell
5205 user -6 0 9008K 9712K pipewr 1:15 0.00% 0.00% ispell
5239 user 10 0 1896K 0K wait 0:00 0.00% 0.00% <sqwebmaild>
5204 user 10 0 1896K 0K wait 0:00 0.00% 0.00% <sqwebmaild>
All the processes are idle, so it's not really an issue.
There is (intentionally) a persistent pool of sqwebmaild processes; you can
change the SQWEBMAIL_PREFORK setting in the sqwebmaild config file if you
want to change the number of processes.
The ispell processes, however, should not be sitting around. Perhaps
sqwebmaild forgot to close its end of the pipes (when they tried to write,
they should have got a SIGPIPE signal). Or perhaps it doesn't leave the
signal handlers in the correct state when it exec's ispell.
Sam?
No, that can't happen. After servicing the request the sqwebmaild process
terminates, taking down all file descriptors with it.