VNXe - How to check Evacuation and Defragmentation


   Article Number:     523359                                   Article Version: 3     Article Type:    How To 




VNXe2 Series,VNXe3200,VNXe1600





There are some processes which runs whenever snapshot/LUN/File System gets into a destroying/deletion state be it manual or automatic.   
    Processes : Below are two prominent processes we have observed in VNXe2 series .   
    1.      Evacuation : Evacuation is the process of releasing the space back to the pool when LUN/File System/Snaps are deleted . It has been observed that this task gets hung or keeps scheduling and cause the Latency , Space issues , High CPU Utilization etc. .   
    2.      Defragmentation : Defragmentation is the process of locating the noncontiguous fragments of data into which a file may be divided as it is stored on physical hard drives . Post which rearranging the fragments and restoring them into fewer fragments or into the whole file.   
    Defragmentation reduces data access time and allows storage to be used more efficiently.   
    Issues : Below are the reasons due to which these Processes gets stuck and causes issues like , Latency , Space Reclamation , High CPU Utilizations .(Snapshot Deletion)   
    Snapshots are created  and if we deleted all snapshots on a LUN completely, that would trigger much more internal metadata IO and block operation in CBFS.   
    This will make the system more busy as more internal IO is generated to the backend.   
    This will cause backend to hit the IO limit per disk. When host IO is coming, the response would be very slow due to these internal IO to backend.   
    The amount of time it takes to delete snapshots varies on object size, system load, number of concurrent deletes, and amount of data that the Snapshot must free back to the container file system when it is destroyed.  It is difficult to predict how long snap deletes will take.      
    High frequency snapshot creation via protection schedules for large LUN, File System, or consistency group objects should be avoided when it is observed that many objects are in the destroying state.   
    Recommendation : If these processes are going on than for the successful progress of these processes , we need to create the snapshot in-order to push these processes and also to reduce the Latency hits , CPU utilization . Evacuation and Defragmentation runs at low priority in background and IOs and other stuffs have high priority.   
    Hence forth , post the snapshot the processes gets speed up .   
    Need to disable the auto creation as well auto deletion of the Snapshot if snapshot is not required in Environment and monitor the Performance .   
    If snapshot creation is required , make sure that every time if any snapshot goes into destroy state another snapshot is created for the same LUN to avoid Performance hits .   
    To check the Evacuation process, please contact support team.