bulletprooftxmanager

type EthBroadcaster ¶

type EthBroadcaster interface {
	Start() error
	Close() error

	Trigger()

	ProcessUnstartedEthTxs(models.Key) error
}

EthBroadcaster monitors eth_txes for transactions that need to be broadcast, assigns nonces and ensures that at least one eth node somewhere has received the transaction successfully.

This does not guarantee delivery! A whole host of other things can subsequently go wrong such as transactions being evicted from the mempool, eth nodes going offline etc. Responsibility for ensuring eventual inclusion into the chain falls on the shoulders of the ethConfirmer.

What ethBroadcaster does guarantee is: - a monotonic series of increasing nonces for eth_txes that can all eventually be confirmed if you retry enough times - transition of eth_txes out of unstarted into either fatal_error or unconfirmed - existence of a saved eth_tx_attempt

func NewEthBroadcaster(store *store.Store, config orm.ConfigReader, eventBroadcaster postgres.EventBroadcaster) EthBroadcaster

type EthResender ¶

type EthResender struct {
	// contains filtered or unexported fields
}

EthResender periodically picks up transactions that have been languishing unconfirmed for a configured amount of time without being sent, and sends their highest priced attempt again. This helps to defend against geth/parity silently dropping txes, or txes being ejected from the mempool.

Previously we relied on the bumper to do this for us implicitly but there can occasionally be problems with this (e.g. abnormally long block times, or if gas bumping is disabled)

func NewEthResender(db *gorm.DB, ethClient eth.Client, pollInterval time.Duration, config orm.ConfigReader) *EthResender

func (er *EthResender) Start()

func (er *EthResender) Stop()

type Log ¶

type Log struct {
	Address     common.Address `json:"address"`
	Topics      []common.Hash  `json:"topics"`
	Data        []byte         `json:"data"`
	BlockNumber uint64         `json:"blockNumber"`
	TxHash      common.Hash    `json:"transactionHash"`
	TxIndex     uint           `json:"transactionIndex"`
	BlockHash   common.Hash    `json:"blockHash"`
	Index       uint           `json:"logIndex"`
	Removed     bool           `json:"removed"`
}

Log represents a contract log event.

Copied from go-ethereum: https://github.com/ethereum/go-ethereum/blob/ce9a289fa48e0d2593c4aaa7e207c8a5dd3eaa8a/core/types/log.go

We use our own version because Geth's version specifies various gencodec:"required" fields which cause unhelpful errors when unmarshalling from an empty JSON object which can happen in the batch fetcher.

func FromGethLog(gl *gethTypes.Log) *Log

func (l Log) MarshalJSON() ([]byte, error)

func (l *Log) UnmarshalJSON(input []byte) error

type NSinserttx ¶

type NSinserttx struct {
	Etx     models.EthTx
	Attempt models.EthTxAttempt
}

NSinserttx represents an EthTx and Attempt to be inserted together

type NonceSyncer ¶

type NonceSyncer struct {
	// contains filtered or unexported fields
}

NonceSyncer manages the delicate task of syncing the local nonce with the chain nonce in case of divergence.

On startup, we check each key for the nonce value on chain and compare it to our local value.

Usually the on-chain nonce will be the same as (or lower than) the next_nonce in the DB, in which case we do nothing.

If we are restoring from a backup however, or another wallet has used the account, the chain nonce might be higher than our local one. In this scenario, we must fastforward the local nonce to match the chain nonce.

The problem with doing this is that now Adamoracle does not have any ownership or control over potentially pending transactions with nonces between our local highest nonce and the chain nonce. If one of those transactions is pushed out of the mempool or re-org'd out of the chain, we run the risk of being stuck with a gap in the nonce sequence that will never be filled.

The solution is to query the chain for our own transactions and take ownership of them by writing them to the database and letting the EthConfirmer handle them as it would any other transaction.

This is not quite as straightforward as one might expect. We cannot query transactions from our account to infinite depth (geth does not support this). The best we can do is to query for all transactions sent within the past ETH_FINALITY_DEPTH blocks and find the ones sent by our address(es).

This gives us re-org protection up to ETH_FINALITY_DEPTH deep in the worst case, which is in line with our other guarantees.

func NewNonceSyncer(store *store.Store, config orm.ConfigReader, ethClient eth.Client) *NonceSyncer

func (s NonceSyncer) SyncAll(ctx context.Context) (merr error)

type Receipt ¶

type Receipt struct {
	PostState         []byte          `json:"root"`
	Status            uint64          `json:"status"`
	CumulativeGasUsed uint64          `json:"cumulativeGasUsed"`
	Bloom             gethTypes.Bloom `json:"logsBloom"`
	Logs              []*Log          `json:"logs"`
	TxHash            common.Hash     `json:"transactionHash"`
	ContractAddress   common.Address  `json:"contractAddress"`
	GasUsed           uint64          `json:"gasUsed"`
	BlockHash         common.Hash     `json:"blockHash,omitempty"`
	BlockNumber       *big.Int        `json:"blockNumber,omitempty"`
	TransactionIndex  uint            `json:"transactionIndex"`
}

Receipt represents an ethereum receipt.

Copied from go-ethereum: https://github.com/ethereum/go-ethereum/blob/ce9a289fa48e0d2593c4aaa7e207c8a5dd3eaa8a/core/types/receipt.go#L50

We use our own version because Geth's version specifies various gencodec:"required" fields which cause unhelpful errors when unmarshalling from an empty JSON object which can happen in the batch fetcher.

func FromGethReceipt(gr *gethTypes.Receipt) *Receipt

func (r Receipt) IsUnmined() bool

func (r Receipt) IsZero() bool

func (r Receipt) MarshalJSON() ([]byte, error)

func (r *Receipt) UnmarshalJSON(input []byte) error