[lustre-discuss] hsm_restore

[Mark Roper]

Lustre Community,

I have a Lustre file system running 2.10.5 using ZFS 0.7.9 as the OST and
MDT file system.  Executing hsm_restore on hundreds of 2 byte files is
causing ~120 MDT writes per second per file. This is with default ZFS
configuration values.

The same hsm_restore on a similarly configured file system using LDISKFS
induces less than 1 write operation per file (presumably because of
batching). As with ZFS, this is using default LDISKFS settings.

Has anyone done ZFS performance tuning or analysis of MDT IOPS with
hsm_restore?

Many thanks!

Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20181113/d078cc83/attachment.html>