Ceph: BlueStore BlueFS Spillover Internals

문제

• Ceph - BlueStore BlueFS Spillover Internals
• RocksDB spillover HDD on compaction

• What are the options slow_total_bytes and slow_used_bytes in BlueStore perf dump?

• From ceph daemon osd.x perf dump command:

  "bluefs": {
      "gift_bytes": 0,
      "reclaim_bytes": 0,
      "db_total_bytes": 32212246528,
      "db_used_bytes": 15892217856,
      "wal_total_bytes": 0,
      "wal_used_bytes": 0,
      "slow_total_bytes": 160009551872,
      "slow_used_bytes": 137363456,
      "num_files": 261,
      "log_bytes": 8749056,
      "log_compactions": 1,
      "logged_bytes": 369836032,
      "files_written_wal": 2,
      "files_written_sst": 607,
      "bytes_written_wal": 12874543619,
      "bytes_written_sst": 36489876338
  },
  

• From ceph osd metadata osd.x command:

  "bluefs": "1",
  "bluefs_db_block_size": "4096",
  "bluefs_db_rotational": "0",
  "bluefs_db_size": "32212254720",
  "bluefs_db_type": "ssd",
  "bluefs_single_shared_device": "0",
  "bluefs_slow_block_size": "4096",
  "bluefs_slow_rotational": "1",
  "bluefs_slow_size": "4000220971008",
  "bluefs_slow_type": "hdd",
  "bluestore_bdev_access_mode": "blk",
  "bluestore_bdev_block_size": "4096",
  "bluestore_bdev_rotational": "1",
  "bluestore_bdev_size": "4000220971008",
  "bluestore_bdev_type": "hdd",