[SOLVED] Ceph high disk usage

[D0peX]

Hi, me again!

Rocking that 5.0 installation now! great work guys.
Although i found a slight issue. using my ceph monitor, it was showing ~240GB usage.
However my pools combined are ~80GB. even with a duplication of 2, that does add up to ~160. not even close to the 240 it is 'thinking' it is using (? if that is how it works, maybe it already count the duplication).



i dont have 'hidden' pools or anything that i know of.

root@hv1:~# ceph osd lspools
0 rbd,1 ceph-vm,2 ceph-lxc,3 scbench,

root@hv1:~# ceph df
GLOBAL:
    SIZE      AVAIL     RAW USED     %RAW USED
    2459G     2223G         236G          9.60
POOLS:
    NAME         ID     USED       %USED     MAX AVAIL     OBJECTS
    rbd          0           0         0          680G           0
    ceph-vm      1      62512M      8.23          680G       15700
    ceph-lxc     2       8930M      1.27          680G        2310
    scbench      3       8880M      1.26          680G        2221
root@hv1:~#

I have no clue how ceph takes all this data in to account, but in my eyes it should be way lower.

Edit during making of the post. i deleted the 'scbench' pool. and this seem to take up 26GB! That pool was 8.67GB? :O, that seems more like a duplication of x3!

i'd love to hear some input / discussion,
D0peX

 

[wolfgang]

Hi,

you have replica 3 not 2 so 240 GB is the expected value.

By the way it is very dangerous to run a pool with replica 3/1.

 

[D0peX]

Ah, i thought that 3/2 meant i have 3 nodes i want to use, and 2x duplication. Could i reduce this number to save some space?

 

[tom]

3/2 is the very recommended setup.

There is a long discussion about this in the Ceph community (mailing lists) if you want to read more.

 

[Belokan]

Ah, i thought that 3/2 meant i have 3 nodes i want to use, and 2x duplication. Could i reduce this number to save some space?

3/2 means 3 replicas and minimum 2 replicas have to be online to make the pool available.

 

[D0peX]

Ah that explains. Sorry i feel dumb now. thank you for your replies!