So we wasted entirely too much time trying to figure out a really complex to diagnose bug at work. Turns out that the startup script for NTP v4 hangs hard if there is no networking enabled when ntp is started on an openSUSE 11 system. Because we run our openSUSE servers in a Xen virtual environment over VLANs, and we don’t bother telling SUSE what ipconfig it should have because Xen fires up the VLAN configs for the O/S; we end up with our dom0 SUSE servers hanging on boot. No error, no warning, no indicator that it was NTP that was part of the problem (brute force sequential chkconfig service off for each service is how we found it).
For the moment we have ntp set to off, then we manually start ntp once the system is up. We’ll do some VLAN bridging and default IP config later so this does not happen again, but wow – ugly. Very happy it is resolved.
Yes, it is a bug. Anything in the startup scripts should not hang hard; properly configured the service will fail and continue allowing the system to boot.
{ 2 } Comments