cephfs directly consuming ec pool

["Markus Hickel" <m.hickel.bg20@xxxxxx>]

Dear all,

I am currently running cephfs directly consuming erasure coded pools on a small test system (4 hosts, ceph osd tree below).
I followed the steps mentioned here: http://ceph.com/community/new-luminous-erasure-coding-rbd-cephfs/
I am running luminous 12.2.2 with bluestore.
During fio tests I experience extremely high memory usage, resulting in mons running oom. With replicated pool only, I experience no problems.
Is this a known behaviour ? If you need additional info, let me know.

 

hosts have dual core cpus and 4 GB memory each.

 

ec profile:
k=2
m=1
 

ceph -s
    health: HEALTH_OK
 
  services:
    mon: 3 daemons, quorum ctest01,ctest02,ctest03
    mgr: ctest02(active), standbys: ctest01
    mds: cephfs-1/1/1 up  {0=ctest02=up:active}
    osd: 8 osds: 8 up, 8 in
 
  data:
    pools:   4 pools, 384 pgs
    objects: 22 objects, 3383 kB
    usage:   13896 MB used, 1830 GB / 1843 GB avail
    pgs:     384 active+clean

 

ceph fs ls:
name: cephfs, metadata pool: cephfs_metadata, data pools: [replicated ec21pool ]

 

ceph osd tree:

ID CLASS WEIGHT  TYPE NAME        STATUS REWEIGHT PRI-AFF
-1       1.80034 root default                             
-9       0.56378     host ctest01                         
 1   hdd 0.29099         osd.1        up  1.00000 1.00000
 7   hdd 0.27280         osd.7        up  1.00000 1.00000
-7       0.72749     host ctest02                         
 0   hdd 0.27280         osd.0        up  1.00000 1.00000
 2   hdd 0.45470         osd.2        up  1.00000 1.00000
-6       0.36368     host ctest03                         
 3   hdd 0.29099         osd.3        up  1.00000 1.00000
 8   hdd 0.07269         osd.8        up  1.00000 1.00000
-8       0.14539     host ctest04                         
 4   hdd 0.07269         osd.4        up  1.00000 1.00000
 5   hdd 0.07269         osd.5        up  1.00000 1.00000

 

Cheers,

Markus

_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com