This function means that the RAID software layer, where redundancy exists to do so, can prevent drive bad behavior from having an unlimited impact on I/O performance. In addition, the system tries to avoid immediately committing to an array rebuild due to a brief offline event from a single drive, while there is full redundancy.
The system temporarily reduces the array redundancy when drive latency impacts a write task. There is a task that is associated with the repair operation to catch up the out of sync component. Other background tasks stop running when drives are slow to lower the workload and expedite recovery.
If the preference is that write I/O does not cause loss of sync, then the CLI supports this function that is disabled on an array-by-array basis through the slowwritepriority parameter to the charray and mkarray commands. This parameter does not affect the RAID layer efforts to return reads promptly when drives misbehave. The slowwritepriority parameter also does not prevent the array from attempting to avoid a full component rebuild by using an incremental rebuild when drives go offline (also displayed as a resync task).
If there is excessive use of the function due to I/O, then the availability of the data can be at risk. To warn of an excessive loss of sync, the array logs an event to help identify if a particular member is the cause. The system monitors drives and fails a drive that is consistently causing slow events.