Comment 3 for bug 47645

Revision history for this message
Dave Williams (dwilliams-cingo) wrote :

Other things I have noticed on the failed install. I am willing to help with these issues.

ccs, cman, clvm all try to load upon installation which means that a working /etc/cluster/cluster.conf must already exist. Suggest either or both:
 - have the installation create a miminal cluster.conf that points to localhost (or something) that prevents these packages from appearing to fail up on installation.
 - install the cluster packages without starting them. The install depends on a kernel update anyway - have the services start on next reboot with the new kernel

Should install modify /etc/lvm/lvm.conf for: locking_library, locking_type, locking_dir

gulm really is depreciated so install should not try to start the service - even on reboot. Recommend removing it from the suite - at least force legacy users to manually configure/start.

some fence scripts that should be added.

sample cluster.conf files for simple two-node and multi-node clusters.