Uploaded image for project: 'Lustre'
  1. Lustre
  2. LU-18791

report overflowed nlink counter correctly for striped dirs.

Details

    • Bug
    • Resolution: Fixed
    • Minor
    • Lustre 2.17.0
    • None
    • 3
    • 9223372036854775807

    Description

      on ldiskfs backend, if shards of a striped dir have more than ~64k subdirectories , the reported nlink attribute for the striped directory is incorrect:

      [root@rocky tests]# ../utils/lfs mkdir -c 2 /mnt/lustre/dir-c2
      [root@rocky tests]# for x in {0..160}; do for y in {0..999} ; do mkdir /mnt/lustre/dir-c2/$x-$y; done; echo .. ${x}k ..; done
      .. 0k ..
      .. 1k ..
      .. 2k ..
      .. 3k ..
      .. 4k ..
      ...
      .. 160k ..
      [root@rocky tests]# stat /mnt/lustre/dir-c2
        File: /mnt/lustre/dir-c2
        Size: 7815168   	Blocks: 15280      IO Block: 1048576 directory
      Device: 2c54f966h/743766374d	Inode: 144115205289279489  Links: 0
      Access: (0755/drwxr-xr-x)  Uid: (    0/    root)   Gid: (    0/    root)
      Access: 2025-03-10 19:50:11.000000000 +0300
      Modify: 2025-03-10 19:53:07.000000000 +0300
      Change: 2025-03-10 19:53:07.000000000 +0300
       Birth: 2025-03-10 19:50:11.000000000 +0300
      [root@rocky tests]# 
      

      as you see, the nlink count is 0.

      Attachments

        Activity

          People

            zam Alexander Zarochentsev
            zam Alexander Zarochentsev
            Votes:
            0 Vote for this issue
            Watchers:
            5 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: