I don't know why, but right now my server started complaining with:
[warn] (128)Network is unreachable: connect to listener
I just issued a "restart" and before it was working fine... Now, even if I
shut it down and start it back again, the error_log is full with those...
I believe it comes out from line 480 in mpm_common.c, but have no clue of
what that means... Weird thing, though, is that httpd works allright (so it
must be only one of the processes, or some of them, not all...
Here's the relevant part in httpd.conf:
ThreadLimit 256
ServerLimit 8
StartServers 2
MaxClients 2048
MinSpareThreads 1
MaxSpareThreads 256
ThreadsPerChild 256
MaxRequestsPerChild 0
If I modify it to be...
ThreadLimit 2048
ServerLimit 1
StartServers 1
MaxClients 2048
MinSpareThreads 1
MaxSpareThreads 2048
ThreadsPerChild 2048
MaxRequestsPerChild 0
(so one process with 2048 threads instead of 8 processes of 256 threads
each) the whole problem zips away... So, it seems that ONLY the first
process is able to bind to the network with WORKER on Solaris 8...
BTW, I noticed one time a file called something like "access.sock", or
really close to that in my logs directory, got created and then destroyed in
no time (and it wasn't CGISOCK! :) :) :)
Do we use by any chance a socket file on disk that, once has been opened,
gets deleted preventing the 2nd, 3rd, 4th and so on processes to bind to it?
(Dumb question, I know, but it might be!) :)
Pier
