none
Problems on deleting big files from CSV volumes RRS feed

  • Question

  • I have a 2 HCI cluster nodes, volume are CSVFS_ReFS, Im using Disk Speed to test storage performance, the results of the tests are ok, but when I delete the file created by the test (around 500GB) the volume hangs, and it takes 15 minutes to the volume become responsive again. The problem does not happen with small files.

    Any ideia on what is causing this issue.


    Monday, December 2, 2019 3:13 PM

Answers

  • I think I found the cause of the problem. I was monitoring the CSV volume through perfmon during the test and I saw that once the test finishes it still had IO on the volume, so I waited the IO stop for a few minutes, and then delete the file, it took only a few seconds to delete the 500GB file.

    I think it is due the cache flhushing data to the volume.

    Thank you for your help.

    Wednesday, December 4, 2019 2:53 PM

All replies

  • Hi ,

    Please remove the CSV and then run chkdsk /scan on the disk to check corrupted file system. Make sure your VMs that sit on the affected CSV are turned off.

    In addition ,disable any anti virus programs to do a test.

    Best Regards,

    Candy


    Please remember to mark the replies as an answers if they help.
    If you have feedback for TechNet Subscriber Support, contact tnmff@microsoft.com   

    Tuesday, December 3, 2019 7:53 AM
  • Hi, thank you for your answer, I will try that as soon as possible. There is no anti virus program on the server, only the default Windows Defender, but I will try to disable it too.
    Tuesday, December 3, 2019 11:39 AM
  • Are you deleting the test file from the node that owns the CSV?  If not, you are dealing with a lot of redirected IO, which will slow things down.  In fact, you also want to run you test from the node that owns the CSV on which the test files will be written if you want the best results.  Of course, the most realistic results will come from production data instead of disk speed.

    tim

    Tuesday, December 3, 2019 1:56 PM
  • Hi Tim, makes sense about the owner node, I will back to my tests in the environment right now, in a few hours I back with answers about this.

    Thanks.

    Tuesday, December 3, 2019 6:54 PM
  • Hi all, I execute the deletion from the owner node, but the time was pretty much the same, 10 to 15 minutes to delete a 500GB file.

    I think that maybe I have something wrong on my network configurations.

    Wednesday, December 4, 2019 12:14 PM
  • I think I found the cause of the problem. I was monitoring the CSV volume through perfmon during the test and I saw that once the test finishes it still had IO on the volume, so I waited the IO stop for a few minutes, and then delete the file, it took only a few seconds to delete the 500GB file.

    I think it is due the cache flhushing data to the volume.

    Thank you for your help.

    Wednesday, December 4, 2019 2:53 PM