Cause (the user action or circumstances that trigger the bug): Previously, when a node left the cluster, the Pacemaker controller would clear its transient node attributes from the CIB, while the attribute manager would clear them from its database. Consequence (what the user experience is when the bug occurs): Timing issues could cause a variety of problems when one component had finished clearing attributes but the other had not. Commonly, this would present as a node immediately leaving the cluster after rejoining, due to its shutdown attribute still being present from its last shutdown. Also, if the controller crashed on a node, that node's transient node attributes would be wrongly erased from the CIB. Fix (what has changed to fix the bug; do not include overly technical details): The attribute manager now handles clearing transient node attribute from both its database and the CIB. Result (what happens now that the patch is applied): Timing issues are no longer possible.