[lustre-discuss] changing inode size on MDT

Mohr Jr, Richard Frank rmohr at utk.edu
Wed Oct 2 13:08:08 PDT 2019



> On Oct 2, 2019, at 3:45 PM, Hebenstreit, Michael <michael.hebenstreit at intel.com> wrote:
> 
> and I'd like to use --mkfsoptions='-i 1024' to have more inodes in the MDT. We already run out of inodes on that FS (probably due to an ZFS bug in early IEEL version) - so I'd like to increase #inodes if possible

I don’t think you can go smaller than “-i 2048”.  I believe the default inode size on newer versions of Lustre is 1KB (to accommodate features like PFL), and Lustre wants an additional 1KB of space per inode for other storage purposes.  I have used “-i 2048” to get more inodes on one of my file systems.  There are a few things you should keep in mind though if you are trying to increase the inode count:

1) You might need to adjust this option for the OSTs as well.  By default, Lustre creates one inode for every 1MB of space on the OST.  If you have a bunch of small files, you might need to increase the number of OST inodes as well.

2) Be aware of things like large PFL layouts that could cause the size of the inode attributes to exceed the default inode size which would cause the MDT to allocate an extra page for storing the attribute info.  That would cause the inode to consume an additional 4KB of space (as I understand it).  A few files like that shouldn’t cause problems, but too many of them could eat up your free MDT space.

—
Rick Mohr
Senior HPC System Administrator
National Institute for Computational Sciences
University of Tennessee





More information about the lustre-discuss mailing list