Batcher configuration

Learn the OP Stack batcher configurations.

This page lists all configuration options for the op-batcher. The op-batcher posts L2 sequencer data to the L1, to make it available for verifiers. The following options are from the --help in v1.10.0.

Batcher policy

The batcher policy defines high-level constraints and responsibilities regarding how L2 data is posted to L1. Below are the standard guidelines for configuring the batcher within the OP Stack.

Parameter	Description	Administrator	Requirement	Notes
Data Availability Type	Specifies whether the batcher uses blobs, calldata, or auto to post transaction data to L1.	Batch submitter address	Ethereum (Blobs or Calldata)	- Alternative data availability (Alt-DA) is not yet supported in the standard configuration. - The sequencer can switch at will between blob transactions and calldata, with no restrictions, because both are fully secured by L1.
Batch Submission Frequency	Determines how frequently the batcher submits aggregated transaction data to L1 (via the batcher transaction).	Batch submitter address	Must target 1,800 L1 blocks (6 hours on Ethereum, assuming 12s L1 block time) or lower	- Batches must be posted before the sequencing window closes (commonly 12 hours by default). - Leave a buffer for L1 network congestion and data size to ensure that each batch is fully committed in a timely manner.

• Data Availability Types:

  • Calldata is generally simpler but can be more expensive on mainnet Ethereum, depending on gas prices.
  • Blobs are typically lower cost when your chain has enough transaction volume to fill large chunks of data.
  • The op-batcher can toggle between these approaches by setting the --data-availability-type=<blobs|calldata|auto> flag or with the OP_BATCHER_DATA_AVAILABILITY_TYPE env variable. Setting this flag to auto will allow the batcher to automatically switch between calldata and blobs based on the current L1 gas price.

• Batch Submission Frequency (OP_BATCHER_MAX_CHANNEL_DURATION and related flags):

  • Standard OP Chains frequently target a maximum channel duration between 1–6 hours.
  • Your chain should never exceed your L2’s sequencing window (commonly 12 hours).
  • If targeting a longer submission window (e.g., 5 or 6 hours), be aware that the safe head can stall up to that duration.

Include these high-level “policy” requirements when you set up or modify your op-batcher configuration. See the batcher configuration reference, which explains each CLI flag and environment variable in depth.

Recommendations

Set your OP_BATCHER_MAX_CHANNEL_DURATION

The default value inside op-batcher, if not specified, is still 0, which means channel duration tracking is disabled. For very low throughput chains, this would mean to fill channels until close to the sequencing window and post the channel to L1 SUB_SAFETY_MARGIN L1 blocks before the sequencing window expires.

To minimize costs, we recommend setting your OP_BATCHER_MAX_CHANNEL_DURATION to target 5 hours, with a value of 1500 L1 blocks. When non-zero, this parameter is the max time (in L1 blocks, which are 12 seconds each) between which batches will be submitted to the L1. If you have this set to 5 for example, then your batcher will send a batch to the L1 every 5*12=60 seconds. When using blobs, because 130kb blobs need to be purchased in full, if your chain doesn’t generate at least ~130kb of data in those 60 seconds, then you’ll be posting only partially full blobs and wasting storage.

• We do not recommend setting any values higher than targeting 5 hours, as batches have to be submitted within the sequencing window which defaults to 12 hours for OP chains, otherwise your chain may experience a 12 hour long chain reorg. 5 hours is the longest length of time we recommend that still sits snugly within that 12 hour window to avoid affecting stability.
• If your chain fills up full blobs of data before the OP_BATCHER_MAX_CHANNEL_DURATION elapses, a batch will be submitted anyways - (e.g. even if the OP Mainnet batcher sets an OP_BATCHER_MAX_CHANNEL_DURATION of 5 hours, it will still be submitting batches every few minutes)

While setting anOP_BATCHER_MAX_CHANNEL_DURATION of 1500 results in the cheapest fees, it also means that your safe head can stall for up to 5 hours.

• This will negatively impact apps on your chain that rely on the safe head for operation. While many apps can likely operate simply by following the unsafe head, often Centralized Exchanges or third party bridges wait until transactions are marked safe before processing deposits and withdrawal.
• Thus a larger gap between posting batches can result in significant delays in the operation of certain types of high-security applications.

Configure your batcher to use multiple blobs

When there’s blob congestion, running with high blob counts can backfire, because you will have a harder time getting blobs included and then fees will bump, which always means a doubling of the priority fees.

The op-batcher has the capabilities to send multiple blobs per single blob transaction. This is accomplished by the use of multi-frame channels, see the specs for more technical details on channels and frames.

A minimal batcher configuration (with env vars) to enable 6-blob batcher transactions is:

This enables blob transactions and sets the target number of frames to 6, which translates to 6 blobs per transaction. The minimum tip cap and base fee are also lifted to 2 gwei because it is uncertain how easy it will be to get 6-blob transactions included and slightly higher priority fees should help. The resubmission timeout is increased to a few minutes to give more time for inclusion before bumping the fees because current transaction pool implementations require a doubling of fees for blob transaction replacements.

Multi-blob transactions are particularly useful for medium to high-throughput chains, where enough transaction volume exists to fill up 6 blobs in a reasonable amount of time. You can use this calculator for your chain to determine what number of blobs are right for you, and what gas scalar configuration to use. Please also refer to guide on Using Blobs for chain operators.

Set your --batch-type=1 to use span batches

Span batches reduce the overhead of OP Stack chains, introduced in the Delta network upgrade. This is beneficial for sparse and low-throughput OP Stack chains.

The overhead is reduced by representing a span of consecutive L2 blocks in a more efficient manner, while preserving the same consistency checks as regular batch data.

Batcher sequencer throttling

This feature is a batcher-driven sequencer-throttling control loop. This is to avoid sudden spikes in L1 DA-usage consuming too much available gas and causing a backlog in batcher transactions. The batcher can throttle the sequencer’s data throughput instantly when it sees too much batcher data built up.

There are two throttling knobs:

1. Transaction throttling, which skips individual transactions whose estimated compressed L1 DA usage goes over a certain threshold, and
2. Block throttling, which caps a block’s estimated total L1 DA usage and leads to not including transactions during block building that would move the block’s L1 DA usage past a certain threshold.

Feature requirements

• This new feature is enabled by default and requires running op-geth version v1.101411.1 or later. It can be disabled by setting --throttle-threshold to 0. The sequencer’s op-geth node has to be updated first, before updating the batcher, so that the new required RPC is available at the time of the batcher restart.
• It is required to upgrade to op-conductor/v0.2.0 if you are using conductor’s leader-aware rpc proxy feature. This conductor release includes support for the miner_setMaxDASize op-geth rpc proxy.

Configuration

Note that this feature requires the batcher to correctly follow the sequencer at all times, or it would set throttling parameters on a non-sequencer EL client. That means, active sequencer follow mode has to be enabled correctly by listing all the possible sequencers in the L2 rollup and EL endpoint flags.

The batcher is configured for throttling by default with the parameters described below (and corresponding flags which allow the defaults to be overridden):

• Backlog of pending block bytes beyond which the batcher will enable throttling on the sequencer via --throttle.unsafe-da-bytes-lower/upper-threshold (env var OP_BATCHER_THROTTLE_UNSAFE_DA_BYTES_LOWER/UPPER_THRESHOLD): 3_200_000 and 12_800_000 (batcher backlog of 3.2MB to 12.8MB of data to batch). Disable throttling by setting the lower threshold to 0. The upper threshold sets the level where the maximum throttling intensity is reached.
• Individual tx size throttling via --throttle.tx-size-lower/upper-limit (env var OP_BATCHER_THROTTLE_TX_SIZE_LOWER/UPPER-LIMIT): 150 and 20_000. This is the limit on the estimated compressed size of a transaction when throttling is at minimum/maximum intensity respectively.
• Block size throttling via --throttle.block-size-lower/upper-limit (env var OP_BATCHER_THROTTLE_BLOCK_SIZE_LOWER/UPPER-LIMIT): 2_000 and 130_000. This is the limit on the estimated compressed size of a block when throttling is at minimum/maximum intensity respectively.
• Throttler controller type via --throttle.controller-type (env var OP_BATCHER_THROTTLE_CONTROLLER_TYPE): quadratic by default. This determines how transaction and block size limits are interpolated when throttling is at intermediate intensity.

For full details, see the readme.

If the batcher at startup has throttling enabled and the sequencer’s op-geth instance to which it’s talking doesn’t have the miner_setMaxDASize RPC enabled, it will fail with an error message like:

lvl=warn msg="Served miner_setMaxDASize" reqid=1 duration=11.22µs err="the method miner_setMaxDASize does not exist/is not available"
In this case, make sure the miner API namespace is enabled for the correct transport protocol (HTTP or WS), see next paragraph.

The new RPC miner_setMaxDASize is available in op-geth since v1.101411.1. It has to be enabled by adding the miner namespace to the correct API flags, like

GETH_HTTP_API: web3,debug,eth,txpool,net,miner
GETH_WS_API: debug,eth,txpool,net,miner

It is recommended to add it to both, HTTP and WS.

Example configuration

This is a basic example of a batcher configuration. Optimal batcher configuration is going to differ for each chain, however you can see some of the most important variables configured below:

  OP_BATCHER_WAIT_NODE_SYNC: true
  OP_BATCHER_CHECK_RECENT_TXS_DEPTH: 5
  OP_BATCHER_POLL_INTERVAL: "5s"
  OP_BATCHER_BATCH_TYPE: "1" # span
  OP_BATCHER_COMPRESSION_ALGO: brotli-10
  OP_BATCHER_DATA_AVAILABILITY_TYPE: auto
  OP_BATCHER_MAX_CHANNEL_DURATION: "150" # up to 30 min to fill blobs
  OP_BATCHER_TARGET_NUM_FRAMES: "5" # 5 blobs, can go to 6 with Pectra activated on L1
  OP_BATCHER_SUB_SAFETY_MARGIN: "300" # 1h safety margin to prevent seq window elapse
  OP_BATCHER_NUM_CONFIRMATIONS: "4"
  OP_BATCHER_NETWORK_TIMEOUT: "10s"
  OP_BATCHER_TXMGR_MIN_BASEFEE: "2.0"
  OP_BATCHER_TXMGR_MIN_TIP_CAP: "2.0"
  OP_BATCHER_TXMGR_FEE_LIMIT_MULTIPLIER: 16 # allow up to 4 doublings
  OP_BATCHER_MAX_PENDING_TX: "10"
  OP_BATCHER_RESUBMISSION_TIMEOUT: "180s" # wait 3 min before bumping fees
  OP_BATCHER_ACTIVE_SEQUENCER_CHECK_DURATION: 5s

Lower throughput chains, which aren’t filling up channels before the MAX_CHANNEL_DURATION is hit, may save gas by increasing the MAX_CHANNEL_DURATION. See the recommendations section.

All configuration variables

Batch generation

batch-type

The batch type. 0 for SingularBatch and 1 for SpanBatch. The default value is 0 for SingularBatch. See the span batch feature page to learn more.

Syntax--batch-type=<value>

max-blocks-per-span-batch

Indicates how many blocks back the batcher should look during startup for a recent batch tx on L1. This can speed up waiting for node sync. It should be set to the verifier confirmation depth of the sequencer (e.g. 4). The default value is 0.

Syntax--check-recent-txs-depth=<value>

compression-algo

The compression algorithm to use. Valid options: zlib, brotli, brotli-9, brotli-10, brotli-11. The default value is zlib.

Syntax--compression-algo=<value>

compressor

The type of compressor. Valid options: none, ratio, shadow. The default value is shadow.

Syntax--compressor=<value>

approx-compr-ratio

Only relevant for ratio compressor. The approximate compression ratio (<=1.0). The default value is 0.6.

Syntax--approx-compr-ratio=<value>

poll-interval

How frequently to poll L2 for new blocks. The default value is 6s.

Syntax--poll-interval=<value>

Batch submission

data-availability-type

Setting this flag to auto will allow the batcher to automatically switch between calldata and blobs based on the current L1 gas price.

The data availability type to use for submitting batches to the L1. Valid options: calldata, blobs, and auto. The default value is calldata.

Syntax--data-availability-type=<value>

target-num-frames

The target number of frames to create per channel. Controls number of blobs per blob tx, if using Blob DA. The default value is 1.

Syntax--target-num-frames=<value>

max-channel-duration

The maximum duration of L1-blocks to keep a channel open. 0 to disable. The default value is 0.

Syntax--max-channel-duration=<value>

sub-safety-margin

The batcher tx submission safety margin (in #L1-blocks) to subtract from a channel’s timeout and sequencing window, to guarantee safe inclusion of a channel on L1. The default value is 10.

Syntax--sub-safety-margin=<value>

max-l1-tx-size-bytes

The maximum size of a batch tx submitted to L1. Ignored for blobs, where max blob size will be used. The default value is 120000.

Syntax--max-l1-tx-size-bytes=<value>

Batcher startup

wait-node-sync

Indicates if, during startup, the batcher should wait for a recent batcher tx on L1 to finalize (via more block confirmations). This should help avoid duplicate batcher txs. The default value is false.

Syntax--wait-node-sync=<value>

check-recent-txs-depth

Indicates how many blocks back the batcher should look during startup for a recent batch tx on L1. This can speed up waiting for node sync. It should be set to the verifier confirmation depth of the sequencer (e.g. 4). The default value is 0.

Syntax--check-recent-txs-depth=<value>

stopped

Initialize the batcher in a stopped state. The batcher can be started using the admin_startBatcher RPC. The default value is false.

Syntax--stopped=<value>

Throttling

The batcher supports throttling configuration to limit DA usage for transactions and blocks, and to control throttle intensity via several controller types (including an experimental PID controller). The following flags are available:

throttle.additional-endpoints

Comma-separated list of endpoints to distribute throttling configuration to (in addition to the L2 endpoints specified with --l2-eth-rpc).

Syntax--throttle.additional-endpoints=<endpoint1>,<endpoint2>,...

throttle.tx-size-lower-limit

The DA size limit for individual transactions when the throttle intensity is at its maximum. 0 disables limits; use 1 as the smallest effective limit.

Syntax--throttle.tx-size-lower-limit=<bytes>

throttle.tx-size-upper-limit

The DA size limit for individual transactions as throttle intensity approaches 0 (minimal throttling). Not applied when throttling is inactive.

Syntax--throttle.tx-size-upper-limit=<bytes>

throttle.block-size-lower-limit

The DA size limit for blocks when throttle intensity is at maximum (applies for linear and quadratic controllers). 0 disables limits; use 1 as the smallest effective limit.

Syntax--throttle.block-size-lower-limit=<bytes>

throttle.block-size-upper-limit

The DA size limit for blocks when throttling is inactive (throttle intensity = 0).

Syntax--throttle.block-size-upper-limit=<bytes>

throttle.controller-type

Type of throttle controller to use. Valid values: step, linear, or quadratic (default).

Syntax--throttle.controller-type=<step|linear|quadratic|pid>

throttle.unsafe-da-bytes-lower-threshold

Threshold on unsafe_da_bytes beyond which the batcher will start to throttle the block builder. 0 disables throttling.

Syntax--throttle.unsafe-da-bytes-lower-threshold=<bytes>

throttle.unsafe-da-bytes-upper-threshold

Threshold on unsafe_da_bytes at which throttling reaches maximum intensity (applies for linear and quadratic controllers).

Syntax--throttle.unsafe-da-bytes-upper-threshold=<bytes>

Transaction manager

num-confirmations

Number of confirmations which we will wait after sending a transaction. The default value is 10.

Syntax--num-confirmations=<value>

max-pending-tx

The maximum number of pending transactions. 0 for no limit. The default value is 1.

Syntax--max-pending-tx=<value>

resubmission-timeout

Duration we will wait before resubmitting a transaction to L1. The default value is 48s.

Syntax--resubmission-timeout=<value>

safe-abort-nonce-too-low-count

Number of ErrNonceTooLow observations required to give up on a tx at a particular nonce without receiving confirmation. The default value is 3.

Syntax--safe-abort-nonce-too-low-count=<value>

txmgr.min-basefee

Enforces a minimum base fee (in GWei) to assume when determining tx fees. 1 GWei by default. The default value is 1.

Syntax--txmgr.min-basefee=<value>

txmgr.min-tip-cap

Enforces a minimum tip cap (in GWei) to use when determining tx fees. 1 GWei by default. The default value is 1.

Syntax--txmgr.min-tip-cap=<value>

fee-limit-multiplier

The multiplier applied to fee suggestions to put a hard limit on fee increases. The default value is 5.

Syntax--fee-limit-multiplier=<value>

txmgr.fee-limit-threshold

The minimum threshold (in GWei) at which fee bumping starts to be capped. Allows arbitrary fee bumps below this threshold. The default value is 100.

Syntax--txmgr.fee-limit-threshold=<value>

txmgr.receipt-query-interval

Frequency to poll for receipts. The default value is 12s.

Syntax--txmgr.receipt-query-interval=<value>

txmgr.not-in-mempool-timeout

Timeout for aborting a tx send if the tx does not make it to the mempool. The default value is 2m0s.

Syntax--txmgr.not-in-mempool-timeout=<value>

txmgr.send-timeout

Timeout for sending transactions. If 0 it is disabled. The default value is 0s.

Syntax--txmgr.send-timeout=<value>

Authentication and wallet

private-key

The private key to use with the service. Must not be used with mnemonic.

Syntax--private-key=<value>

mnemonic

The mnemonic used to derive the wallets for either the service.

Syntax--mnemonic=<value>

hd-path

The HD path used to derive the sequencer wallet from the mnemonic. The mnemonic flag must also be set.

Syntax--hd-path=<value>

signer.address

Address the signer is signing transactions for.

Syntax--signer.address=<value>

signer.endpoint

Signer endpoint the client will connect to.

Syntax--signer.endpoint=<value>

signer.header

Headers to pass to the remote signer. Format key=value.
Value can contain any character allowed in an HTTP header.
When using env vars, split multiple headers with commas.
When using flags, provide one key-value pair per flag.

Syntax--signer.header=<key=value>

signer.tls.ca

tls ca cert path. The default value is tls/ca.crt.

Syntax--signer.tls.ca=<value>

signer.tls.cert

tls cert path. The default value is tls/tls.crt.

Syntax--signer.tls.cert=<value>

signer.tls.key

tls key. The default value is tls/tls.key.

Syntax--signer.tls.key=<value>

Network connections

l1-eth-rpc

HTTP provider URL for L1.

Syntax--l1-eth-rpc=<value>

l2-eth-rpc

HTTP provider URL for L2 execution engine. A comma-separated list enables the active L2 endpoint provider. Such a list needs to match the number of rollup-rpcs provided.

Syntax--l2-eth-rpc=<value>

rollup-rpc

HTTP provider URL for Rollup node. A comma-separated list enables the active L2 endpoint provider. Such a list needs to match the number of l2-eth-rpcs provided.

Syntax--rollup-rpc=<value>

rpc.addr

rpc listening address. The default value is 0.0.0.0.

Syntax--rpc.addr=<value>

rpc.enable-admin

Enable the admin API. The default value is false.

Syntax--rpc.enable-admin=<value>

rpc.port

rpc listening port. The default value is 8545.

Syntax--rpc.port=<value>

network-timeout

Timeout for all network operations. The default value is 10s.

Syntax--network-timeout=<value>

Alt-DA mode (Beta feature)

altda.enabled

Enable Alt-DA mode, Alt-DA mode is a Beta feature of the OP Stack. While it has received initial review from core contributors, it is still undergoing testing, and may have bugs or other issues. The default value is false.

Syntax--altda.enabled=<value>

altda.da-server

HTTP address of a DA Server.

Syntax--altda.da-server=<value>

altda.da-service

Use DA service type where commitments are generated by Alt-DA server. The default value is false.

Syntax--altda.da-service=<value>

altda.get-timeout

Timeout for get requests. 0 means no timeout.

Syntax--altda.get-timeout=<value>

altda.put-timeout

Timeout for put requests. 0 means no timeout.

Syntax--altda.put-timeout=<value>

altda.max-concurrent-da-requests

Maximum number of concurrent requests to the DA server.

Syntax--altda.max-concurrent-da-requests=<value>

altda.verify-on-read

Verify input data matches the commitments from the DA storage service.

Syntax--altda.verify-on-read=<value>

Logging and monitoring

log.level

The lowest log level that will be output. The default value is INFO.

Syntax--log.level=<value>

log.format

Format the log output. Supported formats: ‘text’, ‘terminal’, ‘logfmt’, ‘json’, ‘json-pretty’. The default value is text.

Syntax--log.format=<value>

log.color

Color the log output if in terminal mode. The default value is false.

Syntax--log.color=<value>

log.pid

Show PID in the log.

Syntax--log.pid=<value>

metrics.enabled

Enable the metrics server. The default value is false.

Syntax--version=<value>

Miscellaneous

active-sequencer-check-duration

The duration between checks to determine the active sequencer endpoint. The default value is 2m0s.

Syntax--active-sequencer-check-duration=<value>

pprof.addr

pprof listening address. The default value is 0.0.0.0.

Syntax--pprof.addr=<value>

pprof.enabled

Enable the pprof server. The default value is false.

Syntax--pprof.enabled=<value>

pprof.path

pprof file path. If it is a directory, the path is {dir}/{profileType}.prof.

Syntax--pprof.path=<value>

pprof.port

pprof listening port. The default value is 6060.

Syntax--pprof.port=<value>

pprof.type

pprof profile type. One of cpu, heap, goroutine, threadcreate, block, mutex, allocs.

Syntax--pprof.type=<value>

help

Show help. The default value is false.

Syntax--help=<value>

version

Print the version. The default value is false.

Syntax--version=<value>