On 24/12/2019 16.06, Ch Wan wrote:
Hi all. I deploy a ceph cluster with Mimic 13.2.4. There are 26
nodes, 286 osds and 1.4 PiB avail space.
I created nearly 5,000,000,000 objects by ceph-rgw, each object is 4K
size. So there should be 18TB * 3 disk used. But `ceph df detail` output
shows that the RAW USED is 889 TiB
Is this a bug or I missed something?
bluestore_min_alloc_size is 64KiB for HDDs, so each object is taking up
at least that much space. That's 4834552282 objects * 64 KiB * 3 = 864
TiB, plus some overhead for metadata.
If you want to write tiny objects efficiently, you need to reduce
bluestore_min_alloc_size.
--
Hector Martin (hector@xxxxxxxxxxxxxx)
Public Key: https://mrcn.st/pub
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
