Devkly

클레이튼맛 Geth의 데이터 살펴보기

March 31, 2024

시리즈

기본

  1. 이더리움 상태(State)란 무엇일까?

Geth 동기화

  1. Geth는 어떻게 동기화(Sync)할까?
  2. Geth의 Snapshot과 Snap Sync

Geth DB

해당 시리즈는 Geth DB 시리즈는 Klaytn Dev Ambassador Core Development Project 과정 학습 내용을 공유하기 위한 목적으로 작성하였습니다.

  1. 클레이튼맛 Geth의 데이터베이스 얕게 훑어보기
  2. 클레이튼맛 Geth의 데이터 살펴보기

목차

  • 들어가며
  • 가장 간단하게 살펴보기
  • Genesis Setup Data 분석
  • Transaction Broadcast Data 분석
  • 마치며

들어가며

[key, value]

이전 글에서 Klaytn Geth의 데이터는 위와 같은 구조의 반복이라고 설명했다.

해당 글을 통해 실제 저장된 데이터를 살펴보겠다.

모든 과정은 klaytn-geth-db-analysis를 통해 재현 가능하다.

가장 간단하게 살펴보기

db_manager_structure

다시 이전 글에서 살펴봤던 기본 구조를 참고하자.

결국, 어떠한 Key Value가 Disk에 저장된다.

그렇다면 어떤 데이터가 저장될까?


simpledata1

해당 코드는 db_manager.go에서 확인 가능하다

이러한 Key Value 구조는 전체 chaindata에서 반복된다.

simpledata2

위 이미지에서, 파란색 영역의 마지막 숫자(Block Number)가 계속 증가하는 것을 확인 가능하다.


dataadd

datasize

즉 위와 같은 구조의 데이터가 블록이 생길때마다, 트랜잭션이 생길때마다, 계속 쌓인다.


db_manager_structure

다시 코드 구조를 살펴보자.

  • Prefix를 붙여서 Key를 만든다
  • Block, Tx와 같은 복잡한 데이터는 RLP Encoding을 사용해서 저장한다.

Genesis Setup Data 분석

db_extract/0_genesis_setup

이젠 조금 더 복잡하게 코드를 통해 데이터가 어떻게 쌓이는지 분석해본다.

지금부터 내용은 코드 한줄 한줄을 분석했기 때문에 읽기 힘들다. 필요가 없다면 읽지 않는 것을 추천한다.

Commit

Genesis 블록은 위 로직을 통해 만들어진다.


eth.getBlock(0)

> eth.getBlock(0)
{
 
  hash: "0xf484ef9414585f80efa4918c1d43859073556022a6394da54d4e86b5b485cf4a",
  difficulty: "0x1",
  receiptsRoot: "0x56e81f171bcc55a6ff8345e692c0f86e5b48e01b996cadc001622fb5e363b421",
  
  ...

  transactions: [],
  transactionsRoot: "0x56e81f171bcc55a6ff8345e692c0f86e5b48e01b996cadc001622fb5e363b421",
}

359. g.ToBlock(baseStateRoot, db)

block := g.ToBlock(baseStateRoot, db)

위 함수를 통해 chaindata/statetrie가 생성된다.

363. db.WriteTd(block.Hash(), block.NumberU64(), g.BlockScore)

WriteTd

storage/database/schema.go

Key에 Prefix를 붙인다. (이와 같은 Prefix 방식은 계속 반복)

[
  680000000000000000f484ef9414585f80efa4918c1d43859073556022a6394da54d4e86b5b485cf4a74, 
  01
]

headerPrefix:  68
blockNum: 0000000000000000
blockHash: f484ef9414585f80efa4918c1d43859073556022a6394da54d4e86b5b485cf4a74

364. db.WriteBlock(block)

WriteBlock 

[620000000000000000f484ef9414585f80efa4918c1d43859073556022a6394da54d4e86b5b485cf4a, c1c0]

db.WriteBlock(block)
- dbm.WriteBody(block.Hash(), block.NumberU64(), block.Body())
  - [key] blockBodyKey(blockNum, common.HexToHash(genesisBlockHash))
  - [value] rlp.EncodeToBytes(types.Body{Transactions: []*types.Transaction{}})


[48f484ef9414585f80efa4918c1d43859073556022a6394da54d4e86b5b485cf4a, 0000000000000000]

db.WriteBlock(block)
- dbm.WriteHeader(block.Header())
  - db.Put(headerNumberKey(header.hash), common.Int64ToByteBigEndian(header.Number.Uint64()))


[680000000000000000f484ef9414585f80efa4918c1d43859073556022a6394da54d4e86b5b485cf4a, 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]


db.WriteBlock(block)
- dbm.WriteHeader(block.Header())
  - db.Put(headerKey(number, hash), rlp.EncodeToBytes(header))

Prefix Key 구조는 계속 반복된다. body, header 등의 복잡한 자료구조는 RLP Encoding을 사용한다.

Transaction Broadcast Data 분석

db_extract/1_tx_broadcast

> eth.getBlock(2)
{
  hash: "0x41d4ea0e80f6c91f8b6c8d26ee6937882c7bf88626d345a7b7864725ae010167",
  
  ...

  transactions: ["0x95a13a5e23b815e732416e7f94d9958e0d312f1c1da6f893f547540c3b79bf49"],
  transactionsRoot: "0x571ddd2c39d6bb99f4c0057e1cabffb967c0c387c23072d2778a67d2dc4c6daf",
  uncles: []
}
> eth.getBlock(3)
null

2번 블록은 0x95a13a5e23b815e732416e7f94d9958e0d312f1c1da6f893f547540c3b79bf49 Transaction이 포함되어 있다.

DB Entry: Body

[620000000000000000f484ef9414585f80efa4918c1d43859073556022a6394da54d4e86b5b485cf4a, c1c0] 

# [1번 블록] dbm.WriteBody(block.Hash(), block.NumberU64(), block.Body())
+ [6200000000000000018f82d5ac5647dd9696371b3cc558c0c0f9f74b02fdcf8468acfb1c7d096fe8e7, c1c0]
# [2번 블록] dbm.WriteBody(block.Hash(), block.NumberU64(), block.Body())
+ [620000000000000002749d14765aa26772eda90064bcdf00ea2d2cfe328a6a141dde865012bffd2206, f872f870f86e80850ba43b7400825208948c53ffef58a4e677592354eba1f95b0ea777f96888016345785d8a0000808207f3a0b50e1b4cb415ad25430a92fbb4fdee0b67b3d461da64a91ef58ece34acd1e299a0118fc25940a88b993665613aaa52b3a67739a498e80362f05449ecdcf455c87b]

DB Entry: Receipts

[720000000000000000f484ef9414585f80efa4918c1d43859073556022a6394da54d4e86b5b485cf4a, c0]

# [1번 블록] common.Bytes2Hex(blockReceiptsKey(blockNum, blockHash))
+ [7200000000000000018f82d5ac5647dd9696371b3cc558c0c0f9f74b02fdcf8468acfb1c7d096fe8e7, c0]
# [2번 블록] common.Bytes2Hex(blockReceiptsKey(blockNum, blockHash))
+ [720000000000000002749d14765aa26772eda90064bcdf00ea2d2cfe328a6a141dde865012bffd2206, f90141f9013e01b9010000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000a095a13a5e23b815e732416e7f94d9958e0d312f1c1da6f893f547540c3b79bf49940000000000000000000000000000000000000000c0825208]

DB Entry: Txlookup

# db.Put(TxLookupKey(tx.Hash()), data);
+ [6c95a13a5e23b815e732416e7f94d9958e0d312f1c1da6f893f547540c3b79bf49, e3a0749d14765aa26772eda90064bcdf00ea2d2cfe328a6a141dde865012bffd22060280]

DB Entry: Statetrie

트랜잭션이나 Block Reward로 인한, Statetrie의 변경사항은 매 블록마다 DB에 기록되는것이 아니라, In-Memory에 변경사항을 쌓고 blockInterval(=128 block)마다 주기적으로 DB에 Flush한다.

+ [00a0106cfdd4eb1daee9fadc27bcb8504b9dbdcf4bae8877f242bc8e6d19f61c, f83ea03f3b91639cbb77c038e37d8967a68e36d9f3e0bc15b1bac249b2b51a8aa2cc859c01da0195446c3b15f9926687d2c40534fdb3fedd0b8bf430008001c0]
+ [2ffb37f17f34c9b54cd1882d31327570ea3c6b4acd880416a7ea5d81228ec7ef, f83ea03e5b33e27a71f3ac0bd42b16d1eef796f12a617bed91c4a2cf3ce386b91a6f6c9c01da8095446c3b15f9926687d2c40534fdb564eebe0b40e8008001c0]
+ [5f33454c77feb4342d275cda4bbe2bbc48390926707243a0fe00d8b9397a85e3, f8d180a0e8ae816088497c9d6c5cb8578b150dd58e705533999a11204b1fdf3543749be1a02ffb37f17f34c9b54cd1882d31327570ea3c6b4acd880416a7ea5d81228ec7ef808080a097fac60ef017377a7c40b138b0499e1ba5ff884e03c6a4d9525f3159bc7deb5ca05466a9cb1baf6a022bb7c79e4d2d49f71875626c9a2b19a8000246c4d90447e880a000a0106cfdd4eb1daee9fadc27bcb8504b9dbdcf4bae8877f242bc8e6d19f61c8080808080a0a37ed00e9b1ed642d9ec8bcf8d8b3ae078c71aee104337d756dd8b6c4746165280]
+ [7365637572652d6b65792df237bbd177e17e346d457ab92b5222a942017dbd345f28fb4aaa0a89ed4e2251, 8c53ffef58a4e677592354eba1f95b0ea777f968]
+ [a37ed00e9b1ed642d9ec8bcf8d8b3ae078c71aee104337d756dd8b6c47461652, f1a03237bbd177e17e346d457ab92b5222a942017dbd345f28fb4aaa0a89ed4e22518f01cd8088016345785d8a00008001c0]

위 변경사항은 다른 DB Entry처럼 매 블록마다 기록되는것이 아니라, 128 블록의 생성과 함께 DB에 기록되었다.

tx.toAddress(8c53ffef58a4e677592354eba1f95b0ea777f968)의 데이터(preImage)가 포함되어 있는것을 확인 가능하다.

마치며

해당 글에서는 Klaytn Data가 실제로 어떻게 저장되는지를 필요한 내용 중심으로 살펴보았습니다.

반복되는 내용은 생략했습니다. 더 자세한 내용은 klaytn-geth-db-analysis를 참고해주세요.

Juna
Written by@Juna
I love Node.js