I have a small cluster with a single crash map. I use 3 pools one (Openebula
VMs on rbd), cephfs_data and cephfs_metadata for cephfs. Here is my ceph df
RAW STORAGE:
CLASS SIZE AVAIL USED RAW USED %RAW USED
ssd 94 TiB 78 TiB 17 TiB 17 TiB 17.75
TOTAL 94 TiB 78 TiB 17 TiB 17 TiB 17.75
POOLS:
POOL ID STORED OBJECTS USED %USED
MAX AVAIL
cephfs_data 1 3.3 TiB 6.62M 10 TiB 12.36
24 TiB
cephfs_metadata 2 2.1 GiB 447.63k 2.5 GiB 0
24 TiB
one 5 2.2 TiB 598.12k 6.6 TiB 8.42
24 TiB
What confuses me is an even distribution of MAX_AVAIL storage between those
pools. When I mount cephfs on a client host, df -h shows me pool utilization.
28T 3.4T 24T 13%
I also have an old hammer cluster where I see a similar picture for ceph df for
a one crash map (covering rbd, cephfs-data, cephfs-metadata)
GLOBAL:
SIZE AVAIL RAW USED %RAW USED
87053G 31306G 55747G 64.04
POOLS:
NAME ID USED %USED MAX AVAIL OBJECTS
rbd 0 12907G 69.37 5700G 3312474
cephfs-data 12 2873G 33.52 5700G 5859947
cephfs-meta 13 90035k 0 5700G 443961
cloud12g 14 2857G 43.41 3726G 623737
However, df -h on clients show total cluster utilization
86T 55T 31T 65%
It seems that the hammer dynamically changes allocation data between the pools
on the same crush map as needed. Does nautilus do the same? In this case, does
24 TB means actually avaialble space divided 3 (all my pools are set with 3/2
replication)?
Thank you and sorry for the confusion
_______________________________________________
ceph-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
