Patent attributes
Delayed node upgrade in distributed systems is described herein. A method as described herein can include halting first client disconnections associated with a reboot operation of a file system update at a first computing node of a computing cluster in response to the first computing node being added to a delay list; causing, in further response to the first computing node being added to the delay list, second client disconnections associated with the reboot operation of the file system update at a second computing node of a group of second computing nodes of the computing cluster that are distinct from the first computing node; clearing, in response to the file system update being completed at respective ones of the group of second computing nodes, the delay list; and restarting, in response to clearing the delay list, the first client disconnections at the first computing node.