Red Hat Bugzilla – Bug 63680
ntp enable checkbox DOA
Last modified: 2008-05-01 11:38:02 EDT
Oh great, I thought, a checkbox to enable ntp! This will make it easy. The dateconfig checkbox does not seem to work. I get no feedback about any problem, but the system log shows: Apr 17 08:42:01 hardhat ntpd[3074]: ntpd 4.1.0 Wed Sep 5 06:54:30 EDT 2001 (1) Apr 17 08:42:01 hardhat ntpd: ntpd startup succeeded Apr 17 08:42:01 hardhat ntpd[3074]: precision = 18 usec Apr 17 08:42:01 hardhat ntpd[3074]: kernel time discipline status 0040 Apr 17 08:42:01 hardhat ntpd[3074]: getnetnum: "terrapin.csc.ncsu.edu" invalid host number, line ignored Apr 17 08:42:01 hardhat ntpd[3074]: Un-parsable frequency in /etc/ntp/drift Apr 17 08:42:01 hardhat ntpd[3074]: bind() fd 9, family 2, port 123, addr 224.0.1.1, in_classd=1 flags=0 fails: Address already in use Apr 17 08:42:01 hardhat ntpd[3074]: ...multicast address 224.0.1.1 using wildcard socket I've tried all 4 timeservers in dateconfig, and tried disabling the firewqall. The RedHat default firewall is definitely a problem, but not the only problem.
Can you attach your /etc/ntp.conf file? Looks like the file may have gotten messed up.
Closing due to lack of information.
Created attachment 58284 [details] ntp.conf file, as managed by dateconfig
I eventually gave up dateconfig in frustration, and added this to rc.local: echo -n "Synchronizing with time server ntp.cmr.gov... " /usr/sbin/ntpdate -s -b ntp.cmr.gov This demonstrates the problem is not with ntp.cmr.gov.
Strange, I can connect to the four default servers, the internal Red Hat time server, and also ntp.cmr.gov. I don't know what the problem is. You might try upgrading to the dateconfig from 7.3, but I don't know if that will make a difference or not.
It's not a problem with dateconfig, it's a problem with ntpd as shipped by RedHat. Dateconfig fails to notice the lack of a firewall hole AND some other mystery condition preventing ntp from actually functioning. Dateconfig should be making sure that ntpd can work, and checking that it started up OK (offering a non-functional checkbox to the user is bad).
Ok, I will try to notify the user if starting ntp fails for some reason. Detecting the lack of a firewall hole is hard, though. This isn't just a problem with dateconfig either, it affects pretty much every other network service config tool in the distro. We don't have a consistent way of seeing whether the service we are configuring can actually get out to the outside world. I don't see us having a solution for that anytime soon, though.
I'm still at a loss for WHY ntp fails to start up. It's dateconfig's job to warn that ntp is not working. It's ntp's job to work :-).
I'm going to cc: harald on this report since he is our ntp maintainer. Maybe he can tell what is going on with ntp.
Ok, I've changed dateconfig to notify the user if the ntp server can't be contacted. Please open a new bug against ntp if there's still a problem. Thanks for your report.