Red Hat Bugzilla – Bug 44580
ntp started too early
Last modified: 2007-04-18 12:33:41 EDT
The chkconfig line in /etc/rc.d/init.d/ntpd in ntp-4.0.99k-17 starts ntpd well before named. This is a problem when it tries to resolve the host names in /etc/ntp/step-tickers. Even if you change it to start at position 58 instead of position 26, which I've done, you still have a problem -- named may not actually be finished initializing when the ntpd script runs ntpdate, in which case the ntpdate will fail because it can't resolve the host names in /etc/ntp/step-tickers. I'm not sure what, if anything, can be done about this, short of modifying the named script to wait until it's listening (or until some reasonably short timeout value) before exiting.
what about numeric ip adresses?
Numeric IP addresses are a workaround, not a solution. I have no control over or advance warning when one of my step-tickers changes its IP address.
Even though this was reported earlier, marking it duplicate of the same issue reported on RHL71 (as people tend to check out that Product for known bugs). Adjusting the start position is always a balance. You could argue that it is important that the start times of daemons are _right_ in the logs, and that you can't depend on the DNS when starting the servers _anyway_ (test this via plugging out the cable before booting). Nevertheless, I agree that the advantages of postponing ntpd probably outweigh the disadvantages. *** This bug has been marked as a duplicate of 46464 ***