Red Hat Bugzilla – Bug 164469
failure to resolve any NTP host name results in 2 copies of ntpd running
Last modified: 2007-11-30 17:11:10 EST
Description of problem: Right now, resolving (0,1,2).pool.ntp.org requires a TCP DNS query, and some of us, like me, are cursed with buggy routers that can't handle this properly. When all of the hosts fail to resolve, 2 copies of ntpd are running. Version-Release number of selected component (if applicable): ntp-4.2.0.a.20040617-8 How reproducible: always Steps to Reproduce: 1. Put NTP hostnames in /etc/ntp.conf that can't be resolved (for example, the default ones if one's router is buggy). 2. Start ntpd. Actual results: Two copies of ntpd running. Expected results: One copy of ntpd. Additional info: It would help if pool.ntp.org was configured in a more conventional way, since a firmware update, like a BIOS update, isn't likely before hell freezes over. Please pass it on. (And thanks to bug #164426, I can't even avoid using the router as my DNS server, so there's no workaround.)
Sometime within the last 4 hours or so, the configuration for pool.ntp.org was fixed so it responds to ordinary UDP queries again, and my ntpd starts up properly at boot time. Hence, triggering this bug will require replacing all of the *.pool.ntp.org timeservers with ones that don't currently work.
When ntpd is processing the config file and resolving hostnames, it will fork itself. It's probably not very nice but it's the way it works.