The zFS hang detector automatically monitors the current location of the various tasks processing in zFS. At a set interval, the hang detector thread wakes up and scans the current user requests that have been called into zFS. The hang detector processes this list of tasks and notes various pieces of information to determine the location of the task. When the hang detector determines that a task has remained in the same location for a predefined period of time, it attempts to determine why it is not making progress. This might cause zFS messages or dumps. Certain zFS messages can remain on the screen while the delay continues. If subsequently, the hang detector recognizes that this task has finally progressed, it DOMs the zFS message (removes it from the console). If the zFS message is removed, it means that the delay has cleared and was just a slowdown because of a stressful workload or some other issue. In this case, you can discard any zFS dumps that occur because of this delay.

Several zFS messages warn of potential problems in the zFS address space that have to do with delays. If zFS determines there is a true deadlock, zFS initiates dumps of all systems. The system that detected the deadlock stops and restarts zFS to clear the deadlock. Some delays involve only a single system; other delays in a shared file system environment can involve other systems and XCF communications.