Red Hat Bugzilla – Bug 441658
Fix NM to start early enough so other things can work
Last modified: 2014-03-16 23:13:44 EDT
Description of problem:
During boot, the network init script is not called so eth0 is still down when
the system tries to start ypbind. Enabling the network init via chkconfig and
rebooting solves this problem.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Kickstart install current rawhide tree with NIS enabled
After boot, ypbind is not running even though the network is up. Inspection of
the log file shows that eth0 had not been inited at the time when ypbind was
started (it was started later by NetworkManager). Running "chkconfig network on"
and rebooting solves this problem.
The network is enabled early enough in the boot process for ypbind to run.
OK, with the combination of:
NetworkManager *should* be started earlier for apps to use. Moreover, loopback
is now brought up so that apps can bind even if the final network isn't there.
If you need to wait for NM to fully configure DHCP on boot before proceeding,
1) set NETWORKWAIT in /etc/sysconfig/network
2) file a bug about the application in question, so we can get it fixed
(after all, the network can come and go at any time)
Note: when upgrading to the new hal & NM, you may need to run:
/sbin/chkconfig haldaemon resetpriorities
/sbin/chkconfig NetworkManager resetpriorities
to have the new priorities take effect.
And NetworkManagerDispatcher ? NetworkManagerDispatcher must be started before
NetworkManager if you want the /etc/NetworkManager/dispatcher.d scripts to be
executed with NETWORKWAIT
This should be fixed now.