Ceph » RADOS

I created an 8 OSD cluster with 1 EC pool 5+3 and this ceph df detail output.

$ ceph df detail
RAW STORAGE:
    CLASS     SIZE        AVAIL       USED        RAW USED     %RAW USED
    hdd       8.0 TiB     8.0 TiB     8.1 GiB       16 GiB          0.20
    TOTAL     8.0 TiB     8.0 TiB     8.1 GiB       16 GiB          0.20

POOLS:
    POOL       ID     STORED      (DATA)      (OMAP)     OBJECTS     USED       (DATA)     (OMAP)     %USED     MAX AVAIL     QUOTA OBJECTS     QUOTA BYTES     DIRTY     USED COMPR     UNDER COMPR
    ecpool      1     9.8 MiB     9.8 MiB        0 B         100     50 MiB     50 MiB        0 B         0       4.9 TiB     N/A               N/A               100            0 B             0 B

Using a 9 OSD cluster instead with 1 EC pool 5+3 and this ceph df detail output.

$ ceph df detail
RAW STORAGE:
    CLASS     SIZE        AVAIL       USED        RAW USED     %RAW USED
    hdd       9.0 TiB     9.0 TiB     9.1 GiB       18 GiB          0.20
    TOTAL     9.0 TiB     9.0 TiB     9.1 GiB       18 GiB          0.20

POOLS:
    POOL       ID     STORED      (DATA)      (OMAP)     OBJECTS     USED        (DATA)      (OMAP)     %USED     MAX AVAIL     QUOTA OBJECTS     QUOTA BYTES     DIRTY     USED COMPR     UNDER COMPR
    ecpool      1     9.8 MiB     9.8 MiB        0 B         100     9.8 MiB     9.8 MiB        0 B         0       5.6 TiB     N/A               N/A               100            0 B             0 B

I had thought that num_bytes_hit_set_archive might be non-zero, but rather it is probably the difference between allocated and num_bytes.

  bool use_per_pool_stats() const {
    return osd_sum.num_osds == osd_sum.num_per_pool_osds;
  }

  uint64_t get_allocated_data_bytes(bool per_pool) const {
    if (per_pool) {
      return store_stats.allocated;
    } else {
      // legacy mode, use numbers from 'stats'
      return stats.sum.num_bytes + stats.sum.num_bytes_hit_set_archive;
    }
  }