WiredTiger 스토리지 엔진 구조와 캐시

$ jq -r 'select(.msg == "Opening WiredTiger") | .attr.config' /data/mongod.log
create,cache_size=15508M,session_max=33000,eviction=(threads_min=4,threads_max=4),config_base=false,statistics=(fast),log=(enabled=true,remove=true,path=journal,compressor=snappy),builtin_extension_config=(zstd=(compression_level=6)),file_manager=(close_idle_time=600,close_scan_interval=10,close_handle_minimum=2000),statistics_log=(wait=0),json_output=(error,message),verbose=[recovery_progress:1,checkpoint_progress:1,compact_progress:1,backup:0,checkpoint:0,compact:0,evict:0,history_store:0,recovery:0,rts:0,salvage:0,tiered:0,timestamp:0,transaction:0,verify:0,log:0],prefetch=(available=true,default=false),

test> db.orders.insertMany([{_id: 1, item: "apple", qty: 10}, {_id: 2, item: "banana", qty: 20}, {_id: 3, item: "cherry", qty: 30}])
{ acknowledged: true, insertedIds: { '0': 1, '1': 2, '2': 3 } }
test> db.orders.createIndex({item: 1})
item_1

$ ls -l /data/db
total 104
-rw------- 1 mongod mongod    50 Sep 27 21:41 WiredTiger
-rw------- 1 mongod mongod    21 Sep 27 21:41 WiredTiger.lock
-rw------- 1 mongod mongod  1165 Sep 27 21:41 WiredTiger.turtle
-rw------- 1 mongod mongod  4096 Sep 27 21:41 WiredTiger.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 WiredTigerHS.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 _mdb_catalog.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 collection-0-7164268511808313993.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 collection-2-7164268511808313993.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 collection-4-7164268511808313993.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 collection-7-7164268511808313993.wt
drwx------ 2 mongod mongod  4096 Sep 27 21:41 diagnostic.data
-rw------- 1 mongod mongod  4096 Sep 27 21:41 index-1-7164268511808313993.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 index-3-7164268511808313993.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 index-5-7164268511808313993.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 index-6-7164268511808313993.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 index-8-7164268511808313993.wt
-rw------- 1 mongod mongod 20480 Sep 27 21:41 index-9-7164268511808313993.wt
-rw------- 1 mongod mongod  4096 Sep 27 21:41 internal-10-7164268511808313993.wt
drwx------ 2 mongod mongod  4096 Sep 27 21:41 journal
-rw------- 1 mongod mongod     3 Sep 27 21:41 mongod.lock
-rw------- 1 mongod mongod  4096 Sep 27 21:41 sizeStorer.wt
-rw------- 1 mongod mongod   114 Sep 27 21:41 storage.bson
$ ls -l /data/db/journal
total 204800
-rw------- 1 mongod mongod 104857600 Sep 27 21:41 WiredTigerLog.0000000001
-rw------- 1 mongod mongod 104857600 Sep 27 21:41 WiredTigerPreplog.0000000001
$ cat /data/db/WiredTiger
WiredTiger
WiredTiger 11.3.0: (November 16, 2023)
$ bsondump --quiet /data/db/storage.bson
{"storage":{"engine":"wiredTiger","options":{"directoryPerDB":false,"directoryForIndexes":false,"groupCollections":false}}}
$ cat /data/db/mongod.lock; echo; pgrep -x mongod
33

33

test> db.getSiblingDB("admin").aggregate([{$listCatalog: {}}, {$project: {_id: 0, ns: 1, ident: 1, idxIdent: 1}}])
[
  {
    idxIdent: { _id_: 'index-1-7164268511808313993' },
    ns: 'admin.system.version',
    ident: 'collection-0-7164268511808313993'
  },
  {
    idxIdent: { _id_: 'index-3-7164268511808313993' },
    ns: 'local.startup_log',
    ident: 'collection-2-7164268511808313993'
  },
  {
    idxIdent: {
      _id_: 'index-5-7164268511808313993',
      lsidTTLIndex: 'index-6-7164268511808313993'
    },
    ns: 'config.system.sessions',
    ident: 'collection-4-7164268511808313993'
  },
  {
    idxIdent: {
      _id_: 'index-8-7164268511808313993',
      item_1: 'index-9-7164268511808313993'
    },
    ns: 'test.orders',
    ident: 'collection-7-7164268511808313993'
  }
]
test> db.orders.stats().wiredTiger.uri
statistics:table:collection-7-7164268511808313993
test> Object.entries(db.orders.stats({indexDetails: true}).indexDetails).map(([name, d]) => name + " -> " + d.uri)
[
  '_id_ -> statistics:table:index-8-7164268511808313993',
  'item_1 -> statistics:table:index-9-7164268511808313993'
]

test> const pick = (cs) => cs.split(",").filter(kv => /^(type|key_format|value_format|block_compressor|internal_page_max|leaf_page_max|memory_page_max|split_pct|prefix_compression|log)=/.test(kv))
test> pick(db.orders.stats().wiredTiger.creationString)
[
  'block_compressor=snappy',
  'internal_page_max=4KB',
  'key_format=q',
  'leaf_page_max=32KB',
  'log=(enabled=true)',
  'memory_page_max=10m',
  'prefix_compression=false',
  'split_pct=90',
  'type=file',
  'value_format=u'
]
test> pick(db.orders.stats({indexDetails: true}).indexDetails.item_1.creationString)
[
  'block_compressor=',
  'internal_page_max=16k',
  'key_format=u',
  'leaf_page_max=16k',
  'log=(enabled=true)',
  'memory_page_max=5MB',
  'prefix_compression=true',
  'split_pct=90',
  'type=file',
  'value_format=u'
]

$ mongod --dbpath /data/db --shutdown | grep -v '^{'
Killing process with pid: 33
$ wt -h /data/db list | grep '^table:'
table:_mdb_catalog
table:collection-0-7164268511808313993
table:collection-11-7164268511808313993
table:collection-13-7164268511808313993
table:collection-15-7164268511808313993
table:collection-17-7164268511808313993
table:collection-2-7164268511808313993
table:collection-4-7164268511808313993
table:collection-7-7164268511808313993
table:index-1-7164268511808313993
table:index-12-7164268511808313993
table:index-14-7164268511808313993
table:index-16-7164268511808313993
table:index-18-7164268511808313993
table:index-3-7164268511808313993
table:index-5-7164268511808313993
table:index-6-7164268511808313993
table:index-8-7164268511808313993
table:index-9-7164268511808313993
table:sizeStorer

$ wt -h /data/db dump -x table:collection-7-7164268511808313993 | sed -n '/^Data/,$p'
Data
81
27000000105f69640001000000026974656d00060000006170706c650010717479000a00000000
82
28000000105f69640002000000026974656d000700000062616e616e610010717479001400000000
83
28000000105f69640003000000026974656d00070000006368657272790010717479001e00000000
$ wt -h /data/db dump -x table:collection-7-7164268511808313993 | sed -n '/^Data/,$p' | sed -n 3p | xxd -r -p | bsondump --quiet
{"_id":{"$numberInt":"1"},"item":"apple","qty":{"$numberInt":"10"}}

$ wt -h /data/db dump -x table:index-9-7164268511808313993 | sed -n '/^Data/,$p'
Data
3c6170706c6500040008

3c62616e616e6100040010

3c63686572727900040018

$ wt -h /data/db dump -x table:index-8-7164268511808313993 | sed -n '/^Data/,$p'
Data
2b0204
0008
2b0404
0010
2b0604
0018

$ wt -h /data/db verify -d dump_layout table:collection-11-7164268511808313993 2>&1 | grep -vE '^\['
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
file:collection-11-7164268511808313993.wt, ckpt_name: WiredTigerCheckpoint.1
Root:
	> addr: [0: 10199040-10203136, 4096, 3539864155]
Internal page tree-depth (total 4):
	001: 1
	002: 3
Leaf page tree-depth (total 580):
	003: 580
$ wt -h /data/db verify -d dump_layout table:collection-17-7164268511808313993 2>&1 | grep -vE '^\['
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
file:collection-17-7164268511808313993.wt, ckpt_name: WiredTigerCheckpoint.1
Root:
	> addr: [0: 28393472-28397568, 4096, 2100132084]
Internal page tree-depth (total 6):
	001: 1
	002: 5
Leaf page tree-depth (total 991):
	003: 991

$ wt -h /data/db verify -d dump_address table:collection-11-7164268511808313993 2>&1 | head -9 | cut -c1-44
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
file:collection-11-7164268511808313993.wt, c
Root:
	> addr: [0: 10199040-10203136, 4096, 353986
[NoAddr] -/-,-/- row-store internal write ge
[0: 10186752-10190848, 4096, 4021143610] new
[0: 4096-40960, 36864, 1162190748] newest_du
[0: 40960-77824, 36864, 3643813452] newest_d
[0: 77824-114688, 36864, 3315467987] newest_
$ wt -h /data/db verify -d dump_address table:collection-17-7164268511808313993 2>&1 | head -9 | cut -c1-44
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
file:collection-17-7164268511808313993.wt, c
Root:
	> addr: [0: 28393472-28397568, 4096, 210013
[NoAddr] -/-,-/- row-store internal write ge
[0: 28372992-28377088, 4096, 3467847535] new
[0: 4096-32768, 28672, 2770148487] newest_du
[0: 32768-61440, 28672, 1153302453] newest_d
[0: 61440-90112, 28672, 553415368] newest_du

$ wt -h /data/db stat table:collection-11-7164268511808313993 | grep -E 'btree: (maximum (internal|leaf) page size|number of key/value pairs|row-store (internal|leaf) pages)|block-manager: file size'
block-manager: file size in bytes=10M (10211328)
btree: maximum internal page size=4096
btree: maximum leaf page size=32768
btree: number of key/value pairs=100000
btree: row-store internal pages=4
btree: row-store leaf pages=580

test> db.createCollection("c_snappy")
{ ok: 1 }
test> db.createCollection("c_zstd", {storageEngine: {wiredTiger: {configString: "block_compressor=zstd"}}})
{ ok: 1 }
test> db.createCollection("c_zlib", {storageEngine: {wiredTiger: {configString: "block_compressor=zlib"}}})
{ ok: 1 }
test> db.createCollection("c_none", {storageEngine: {wiredTiger: {configString: "block_compressor=none"}}})
{ ok: 1 }
test> ["c_snappy", "c_zstd", "c_zlib", "c_none"].map(c => c + ": " + db[c].stats().wiredTiger.creationString.match(/block_compressor=\w*/)[0])
[
  'c_snappy: block_compressor=snappy',
  'c_zstd: block_compressor=zstd',
  'c_zlib: block_compressor=zlib',
  'c_none: block_compressor=none'
]

test> db.c_snappy.findOne({_id: 7})
{
  _id: 7,
  user: 'user18224',
  email: 'user7@example.com',
  city: 'Seoul',
  status: 'inactive',
  amount: 6594.56,
  tags: [ 'member', 'cart' ],
  createdAt: ISODate('2026-06-16T23:31:02.336Z'),
  memo: 'order order cart sale return cart order sale order order sale order return sale return'
}

test> db.adminCommand({fsync: 1})
{ numFiles: 1, ok: 1 }
test> ["c_snappy", "c_zstd", "c_zlib", "c_none"].map(c => { const s = db[c].stats(); return {coll: c, count: s.count, size: s.size, storageSize: s.storageSize, ratio: (s.size / s.storageSize).toFixed(2)}; })
[
  {
    coll: 'c_snappy',
    count: 100000,
    size: 27461165,
    storageSize: 10211328,
    ratio: '2.69'
  },
  {
    coll: 'c_zstd',
    count: 100000,
    size: 27461165,
    storageSize: 4886528,
    ratio: '5.62'
  },
  {
    coll: 'c_zlib',
    count: 100000,
    size: 27461165,
    storageSize: 4882432,
    ratio: '5.62'
  },
  {
    coll: 'c_none',
    count: 100000,
    size: 27461165,
    storageSize: 28405760,
    ratio: '0.97'
  }
]

$ ls -l /data/db/collection-11-7164268511808313993.wt /data/db/collection-13-7164268511808313993.wt /data/db/collection-17-7164268511808313993.wt
-rw------- 1 mongod mongod 10211328 Sep 27 21:41 /data/db/collection-11-7164268511808313993.wt
-rw------- 1 mongod mongod  4886528 Sep 27 21:41 /data/db/collection-13-7164268511808313993.wt
-rw------- 1 mongod mongod 28405760 Sep 27 21:41 /data/db/collection-17-7164268511808313993.wt

test> function cacheOf(c) { const s = db[c].stats(); return {coll: c, storageSize: s.storageSize, size: s.size, inCache: s.wiredTiger.cache["bytes currently in the cache"], readIntoCache: s.wiredTiger.cache["bytes read into cache"]}; }
[Function: cacheOf]
test> cacheOf("c_snappy")
{
  coll: 'c_snappy',
  storageSize: 10211328,
  size: 27461165,
  inCache: 383,
  readIntoCache: 9059
}
test> db.c_snappy.aggregate([{$group: {_id: null, n: {$sum: 1}, total: {$sum: "$amount"}}}])
[ { _id: null, n: 100000, total: 494173742.55 } ]
test> cacheOf("c_snappy")
{
  coll: 'c_snappy',
  storageSize: 10211328,
  size: 27461165,
  inCache: 31422776,
  readIntoCache: 28220035
}
test> db.c_none.aggregate([{$group: {_id: null, n: {$sum: 1}, total: {$sum: "$amount"}}}])
[ { _id: null, n: 100000, total: 494173742.55 } ]
test> cacheOf("c_none")
{
  coll: 'c_none',
  storageSize: 28405760,
  size: 27461165,
  inCache: 31511577,
  readIntoCache: 28249003
}

// Set a minimum of 256MB, otherwise use 50% of available memory over 1GB.
cacheSizeMB = std::max((memSizeMB - 1024) * 0.5, 256.0);

$ cat /sys/fs/cgroup/memory.max
max
$ jq -r 'select(.msg == "Opening WiredTiger") | .attr.config' /data/mongod.log | grep -oE 'cache_size=[0-9]+M'
cache_size=15508M

test> const h = db.adminCommand({hostInfo: 1}).system; ({memSizeMB: h.memSizeMB, memLimitMB: h.memLimitMB, cacheMaxBytes: db.serverStatus().wiredTiger.cache["maximum bytes configured"]})
{
  memSizeMB: Long('32041'),
  memLimitMB: Long('32041'),
  cacheMaxBytes: Long('16261316608')
}

$ cat /sys/fs/cgroup/memory.max
1073741824
$ jq -r 'select(.msg == "Opening WiredTiger") | .attr.config' /data/mongod.log | grep -oE 'cache_size=[0-9]+M'
cache_size=256M

test> const h = db.adminCommand({hostInfo: 1}).system; ({memSizeMB: h.memSizeMB, memLimitMB: h.memLimitMB, cacheMaxBytes: db.serverStatus().wiredTiger.cache["maximum bytes configured"]})
{
  memSizeMB: Long('32041'),
  memLimitMB: Long('1024'),
  cacheMaxBytes: 268435456
}

// /data/cache.js
function cache() {
  const c = db.serverStatus().wiredTiger.cache, mb = (b) => Math.round(b / 1048576);
  const max = c["maximum bytes configured"];
  return {maxMB: mb(max), inCacheMB: mb(c["bytes currently in the cache"]), usedPct: +(100 * c["bytes currently in the cache"] / max).toFixed(1),
    dirtyMB: mb(c["tracked dirty bytes in the cache"]), dirtyPct: +(100 * c["tracked dirty bytes in the cache"] / max).toFixed(1),
    pagesRead: c["pages read into cache"], pagesWritten: c["pages written from cache"],
    workerEvict: c["evict page attempts by eviction worker threads"], appEvict: c["page evict attempts by application threads"],
    appEvictUsecs: c["application thread time evicting (usecs)"],
    cleanEvicted: c["unmodified pages evicted"], dirtyEvicted: c["modified pages evicted"],
    dirtyTrigger: c["number of times dirty trigger was reached"], evictionTrigger: c["number of times eviction trigger was reached"]};
}

// /data/load-ev.js: 1KB 남짓한 문서 40만 개(약 400MB)를 넣는다
let seed = 7;
const rnd = () => (seed = (seed * 1103515245 + 12345) % 2147483648);
let pool = "";
while (pool.length < 1000000) pool += rnd().toString(36);
for (let b = 0; b < 400; b++) {
  const docs = [];
  for (let i = 0; i < 1000; i++)
    docs.push({_id: b * 1000 + i, user: "user" + (rnd() % 100000), amount: rnd() % 100000, memo: pool.substr(rnd() % 990000, 1000)});
  db.ev.insertMany(docs);
}

test> cache()
{
  maxMB: 256,
  inCacheMB: 0,
  usedPct: 0,
  dirtyMB: 0,
  dirtyPct: 0,
  pagesRead: 23,
  pagesWritten: 3,
  workerEvict: 0,
  appEvict: 0,
  appEvictUsecs: 0,
  cleanEvicted: 0,
  dirtyEvicted: 0,
  dirtyTrigger: 0,
  evictionTrigger: 0
}

$ time mongosh --quiet /data/load-ev.js

real	0m2.046s
user	0m1.521s
sys	0m0.176s

test> cache()
{
  maxMB: 256,
  inCacheMB: 180,
  usedPct: 70.4,
  dirtyMB: 1,
  dirtyPct: 0.5,
  pagesRead: 32,
  pagesWritten: 8979,
  workerEvict: 688,
  appEvict: 63,
  appEvictUsecs: 17973,
  cleanEvicted: 601,
  dirtyEvicted: 248,
  dirtyTrigger: 0,
  evictionTrigger: 0
}
test> const s = db.ev.stats(); ({count: s.count, sizeMB: Math.round(s.size / 1048576), storageSizeMB: Math.round(s.storageSize / 1048576)})
{ count: 400000, sizeMB: 403, storageSizeMB: 336 }

test> let t = Date.now(); db.ev.updateMany({}, {$inc: {amount: 1}}).modifiedCount + " docs, " + (Date.now() - t) + " ms"
400000 docs, 967 ms
test> cache()
{
  maxMB: 256,
  inCacheMB: 206,
  usedPct: 80.6,
  dirtyMB: 11,
  dirtyPct: 4.3,
  pagesRead: 5128,
  pagesWritten: 22401,
  workerEvict: 10952,
  appEvict: 68,
  appEvictUsecs: 18383,
  cleanEvicted: 2395,
  dirtyEvicted: 8770,
  dirtyTrigger: 9,
  evictionTrigger: 0
}

test> t = Date.now(); db.ev.aggregate([{$group: {_id: null, n: {$sum: 1}, total: {$sum: "$amount"}}}]).toArray()[0].n + " docs, " + (Date.now() - t) + " ms"
400000 docs, 143 ms
test> cache()
{
  maxMB: 256,
  inCacheMB: 199,
  usedPct: 77.8,
  dirtyMB: 11,
  dirtyPct: 4.3,
  pagesRead: 15308,
  pagesWritten: 22401,
  workerEvict: 21297,
  appEvict: 72,
  appEvictUsecs: 18388,
  cleanEvicted: 12701,
  dirtyEvicted: 8770,
  dirtyTrigger: 9,
  evictionTrigger: 1
}