by I am running Solaris 2.4 on Sparc 20's and have installed the latest jumbo
patches.
I run 14 copies of the CERN httpd 3.0 (libwww 2.17) server. Every night at
midnight I have a script that sends each server a SIGHUP to rotate the
log files.
Two days ago, I noticed that :
1) the httpd proceses were no longer running as nobody/nogroup, instead they
were root.
2) the httpd processes no longer respond to SIGHUP's. I can't get them to
reload the configuration files nor rotate the log files.
The only thing that has changed is that I had moved the configuration files
to new names (and restarted all the servers). Initially I had the conf files
named "httpd.conf.domain" where domain was the name of the domain that the
server responds to. Given that this becomes pretty long I changed the
names to "domain" in a separate subdirectory and then reran all of the
httpd processes using the new configuration files.
Then SIGHUP's no longer worked. I went back and renamed everything to
httpd.conf.domain, restarted everything and it still didn't work.
This configuration has been working flawlessly for MONTHS.
I send a "kill -HUP <pid>" directly to the process and it still does nothing.
Could it be something strange like persmissions on the configuration files?
Does httpd do some checking there?
Why wouldn't it be changing the UID to nobody/nogroup the way it did before?
Any help or suggestions would be greatly appreciated. Please email
thanks,
-- Yermo
-- Internet Connectivity/Services/Training/Consulting/Development
---------------------------------------------------------------------------
Technical Director and Md, USA
Fearless Systems Administrator http://www.radix.net
RadixNet
Opinions expressed are my own.