[zfs-discuss] "du -ksh ." taking hours to complete

Richard Laager rlaager at wiktel.com
Tue Nov 13 19:17:13 EST 2018

On 11/13/18 6:06 PM, Welison B. Floriano via zfs-discuss wrote:
> We are running Red Hat 7.6, zfs 0.7.11-1, 5x raidz2(12
> x 10TB HDs each). 296TB used, 107TB free. Read and write seems to be
> fine, but when we use "du -ksh ." in directories with lost of files, it
> takes hours to get a result. Any help will be appreciated.

Hopefully you'll get a more in-depth response from someone else, but I
just wanted to start with the basics:

A raidz vdev gives the IOPS of a _single_ disk. So you're getting the
IOPS of 5 disks, not 5*10 = 50 as you might be expecting.

I assume that stat()ing the files in a large directory is causing a lot
of small requests. This, combined with the above, is why things are slow.


More information about the zfs-discuss mailing list