They shouldn't be running by default.
In RHEL, the packages are in a subchannel. In FC, they are in core. This makes the difference between
why they should be on by default or off.
In the case of bug 156922, ccsd "spams" to syslog complaining that it can't connec to the cluster. In
FC, the user may never intend to set up a cluster.
If they shouldn't be running by default, then they probably shouldn't be
installed on the system. A little more detail as to why you think your opinion
is the correct behavior would be helpfull.
The scripts are designed so that once the cluster is configured, a node will
automatically be able to become a part of it without any additional
configuration. I do not see any justification for changing the init script
behavior at this time.
The correct person to pawn this bug off on is either the rpm package
maintainer(s) (where chkcofig --add is being called) or the FC installer maintainer.
I think correct behaviour is to change the chkconfig lines in the init scripts
such that the services do not start by default in any runlevels, i.e.
# chkconfig: - 20 80
whatever tool performs cluster configuration should do "chkconfig ccsd on" or
equivalent to then enable these services in appropriate runlevels on the machine.
This should be fixed in ccs-0.25-0.3 and beyond.
Note: we only turn of the init scripts on new installs (so if you upgrade ccsd
will still attempt to start). You can disable the ccsd init script manually by
running '/sbin/chkconfig ccsd off'.