Nimbus v23.5.1
is a medium-urgency
point release improving the compatibility of Nimbus with 3rd party validator clients and beacon nodes and introducing the support for incremental pruning. If you are still not using the --history:prune
option, we recommend testing it in a non-production environment, as it will be enabled by default in our next release.
-
The Nimbus validator client no longer accepts under-specified beacon node URLs that doesn't include a port number or a protocol scheme. When a protocol scheme is specified, Nimbus now uses the default port for the selected protocol (80 for HTTP and 443 for HTTPS):
-
The history pruning is now incremental and no longer results in start-up delays when the
--history:prune
option is enabled on an existing node: status-im#4887 -
Nimbus now uses the withdrawal address of the validator as a default choice for the fee recipient address if the user has not provided any value in the configuration: status-im#4968
-
Nimbus now supports the upcoming Capella hard-fork in the Gnosis network: status-im#4936
-
The Capella-related properties
MAX_BLS_TO_EXECUTION_CHANGES
,MAX_WITHDRAWALS_PER_PAYLOAD
,MAX_VALIDATORS_PER_WITHDRAWALS_SWEEP
andDOMAIN_BLS_TO_EXECUTION_CHANGE
were missing from the/eth/v1/config/spec
REST API end-point: status-im#4925 -
The
/eth/v1/validator/blinded_blocks/{slot}
was supplying incorrectly encoded response when requested to return SSZ data: status-im#4943 -
The safety checks associated with the
--weak-subjectivity-checkpoint
parameter are now compliant with the latest Ethereum specs: status-im#4923 -
The Nimbus validator client was using HTTP pipelining which is not supported by all beacon node implementations: status-im#4950
-
The "Connection to EL node degraded" warning is now printed only after sufficiently persistent connectivity issues with the EL client: status-im#4960
-
After being only briefly disconnected from the execution layer client, the Nimbus beacon node was prematurely setting the
execution_optimistic
flag when returning validator duties: status-im#4955 -
Nimbus now allows the builder to respond 500ms later than the spec-mandated timeout in order to account for possible additional delays introduced by proxies such as mev-boost: status-im#4964
-
During sync committee period transitions, for a brief period of time there was a low risk of producing an invalid sync committee contribution: status-im#4953
-
Nimbus
v23.5.0
introduced an unintended backwards-incompatible change in the parsing of remote keystores which is addressed in this release: status-im#4967
Nimbus v23.5.0
is a medium-urgency
upgrade that addresses a critical issue which was introduced in the previous version (v23.4.0
). The issue was causing missed block proposals for users who were utilizing an external builder.
-
After Nimbus completes a trusted node sync executed with the
--trusted-block-root
flag, it will enable signature verification of all backfilled blocks, thereby reducing the assumed trust in the specified beacon node URL to merely expected data availability rather than expected data authenticity: -
The
/eth/v1/node/syncing
BeaconAPI endpoint now supports the standardizedel_offline
property: -
The
secp256k1
library has been upgraded to version0.3.1
. -
Nimbus now supports an experimental extension of the Web3Signer protocol, allowing the signer server to verify certain properties of the signed block, such as the specified fee recipient:
https://nimbus.guide/web3signer.html#verifying-web3signer status-im#4775 status-im#4912
-
Nimbus was submitting blocks with incorrect state root to the attached external builder which resulted in missed block proposals:
-
Nimbus was skipping block proposals due to an inappropriate triggering of the slashing protection logic when an external builder was providing a block with insufficient value to be selected under the new
--local-block-value-boost
mechanism: -
Nimbus was crashing after certain unsuccessful requests to the external block builder:
-
The Nimbus validator client was failing to perform sync committee duties when attached to multiple beacon nodes and when some of them were only optimistically synced:
-
The
--trusted-block-root
option was not visible in thetrustedNodeSync
help listing: -
Nimbus was experiencing sporadic request time outs when being connected to the execution client over HTTP. Under specific circumstances this was introducing risk for missed attestation:
https://github.com/status-im/nimbus-eth2/commit/d784672c107f846163082262334d4d7a4e625bd5
-
The required traffic to the execution client was reduced by preventing the sending of the same block multiple times:
Nimbus v23.4.0
is a medium-urgency
upgrade addressing a number of low probability risks for missed block proposals, bringing performance improvements in setups relying on the Nimbus validator client, and introducing some exciting new capabilities of the Nimbus light client and Builder API implementations.
-
Nimbus now obtains blocks from the configured builder and execution layer nodes without providing timing advantage to any source. You can use the newly added
--local-block-value-boost
option to give preference to the best block provided by an execution layer node, as long as its value is within the specified percentage of the value advertised by the best external builder. Setting this flag to a non-zero value is recommended because the usage of an external builder introduces an additional risk that the advertised block won't be published by the builder: -
Nimbus now supports the standardized Builder API liveness failsafe mechanism:
-
The
--sync-light-client
option is now enabled by default, providing significant speedups in beacon chain syncing and re-syncing: -
The
trustedNodeSync
command features a new--trusted-block-root
option that leverages the Nimbus light client in order to minimize the required trust in the specified Beacon API endpoint. After downloading the state snapshot, the light client will verify that it conforms to the established consensus in the network. Note that the provided--trusted-block-root
should be somewhat recent, and that additional security precautions such as comparing the state root against block explorers are still recommended. -
Improved scheduling mechanisms in the Nimbus validator client deliver stability and performance improvements:
-
The
deposits exit
command can now be used to perform voluntary exits for multiple validators at once: -
Nimbus now supports the
/eth/v1/beacon/states/{state_id}/randao
REST API endpoint: -
Nimbus now uses only the Capella-enabled
engine_forkchoiceUpdatedV2
endpoint in all communication with the execution layer:
-
Nimbus has addressed a risk of missed block proposal due to incorrectly computed withdrawals at epoch boundaries:
-
Nimbus has addressed a low probability risk of missed block proposals when the configured builder doesn't respond in time:
-
Nimbus has addressed a low probability risk of missed block proposals when a late block triggers a chain re-org while an
engine_forkchoiceUpdated
request to the execution layer is in flight: -
Nimbus will no longer experience occasional response timeouts when performing a large number of concurrent HTTP requests (e.g. when configured to operate with a large number of remote keystores):
status-im/nim-presto#44 status-im/nim-chronos#324 status-im#4779
-
The Nimbus validator client will no longer crash on start-up when supplied with incorrect beacon node configuration:
-
Nimbus will no longer crash when there is a network mismatch between the imported slashing protection database and the specified data directory:
-
Inactive validators will no longer affect the initial GossipSub topic subscriptions:
-
Failed or timed out request to
engine_exchangeTransitionConfigurationV1
will no longer degrade the status of the connection to the execution layer:
Nimbus v23.3.2
is a low-urgency
, but mandatory upgrade providing full-support for the upcoming Capella hard-fork on Mainnet. Please upgrade at your earliest convenience - before the 12th of April.
-
The
deposits exit
can now be executed with a path to a keystore file that was generated bydeposit-staking-cli
orethdo
. All users are advised to use this method for exiting, due to a known issue preventing the other formerly supported methods from working: -
The metrics
beacon_light_client_finality_update_received
,beacon_light_client_finality_update_dropped
,beacon_light_client_optimistic_update_received
andbeacon_light_client_optimistic_update_dropped
provide information regarding the observed light client gossip traffic: -
Nimbus now recognizes the
/eth/v1/validator/beacon_committee_selections
and/eth/v1/validator/sync_committee_selections
Beacon API end-points in accordance to the latest spec:
-
Nimbus will no longer report warnings such as "Connection to EL node degraded" when paired with an execution node that hasn't been synced up to the deployment block of the validator deposit contract:
-
Nimbus was sporadically triggering an inappropriate assertion error under normal operating conditions:
Nimbus v23.3.1
is a medium-urgency
point release addressing a number of accidental configuration handling breaking changes that were shipped in the v23.3.0
release. It also improves the stability of Nimbus when paired with a Besu execution client and improves the fault-tolerance when driving multiple execution clients.
-
Nimbus was performing
eth_getLogs
request with parameters that were exceeding the default--rpc-max-logs-range=1000
limit on Besu. This was a non-fatal issue that resulted in slower deposit syncing speed and the frequent warning message "Connection to EL node degraded". The limit will be increased in the next mainnet release of Besu, but Nimbusv23.3.1
honours the existing limit at the cost of a slightly slower syncing speed with all other execution clients:https://github.com/status-im/nimbus-eth2/commit/6fb48aca7dedc7ba3c6b2f2ae8a4926ddcf7a00e
-
v23.3.0
did not support Engine API URLs which don't specify a protocol in the URL (e.g.http
,https
,ws
orwss
).v23.3.1
is backwards-compatible with all previous Nimbus releases:https://github.com/status-im/nimbus-eth2/commit/3a35809a02b4fbe23b2dc843806ec81f67521c6d
-
v23.3.0
produced a parsing error on TOML configuration files that specify theweb3-url
parameter as an array of strings.v23.3.1
is backwards-compatible with all previous Nimbus releases and introduces a new more convenient way for specifying the Engine API configuration in TOML:https://nimbus.guide/eth1.html#running-multiple-execution-clients https://github.com/status-im/nimbus-eth2/commit/46f48269ef899f19cd9932b27d30c68e2ccf035b
-
v23.3.0
removed the hidden configuration option--web3-force-polling
which remained in use by some users.v23.3.1
restores the option as a deprecated one. Please note that all hidden configuration options are intended for use only by the Nimbus development team for testing purposes:https://github.com/status-im/nimbus-eth2/commit/ee610cbf34cebea24576c25bf6702de4205a260a
-
The release addresses a potential crash triggered by Engine API connections experiencing frequent error responses:
https://github.com/status-im/nimbus-eth2/commit/d899a6a834c083a62e1246eade5027a7019ace82
-
The release addresses a potential issue where a single non-synced execution client may cause the Nimbus sync state to revert to
synced/opt
, even when all validator duties can be performed through the remaining execution clients that are still synced:https://github.com/status-im/nimbus-eth2/commit/d899a6a834c083a62e1246eade5027a7019ace82
Nimbus v23.3.0
is a low-urgency
upgrade bringing full support for the upcoming Capella hard-fork on the Goerli testnet. Keep an eye out for future mainnet releases!
-
You can increase the resilience of your setup and eliminate any downtime during upgrade procedures of the execution client by allowing your beacon node to manage multiple execution clients. To enable this mode, just specify multiple URLs through the
--el
option (alias of--web3-url
) when starting your beacon node:./run-mainnet-beacon-node.sh \ --el=http://127.0.0.1:8551 \ --el=ws://other:8551 \ --jwt-secret=/tmp/jwtsecret
As long as any of execution clients remains operational and fully synced, Nimbus will keep performing all validator duties. To carry out an upgrade procedure without any downtime, just restart the execution clients one by one, waiting for each instance to re-sync before moving to the next one.
If you use this mode with different execution client implementations, Nimbus will act as an execution layer consensus violation detector, preventing the publishing of blocks that may trigger a catastrophic partitioning in the network.
-
The metrics
engine_api_responses
,engine_api_request_duration_seconds
andengine_api_timeouts
provide statistics about latency and response status codes for all requests sent to each individual execution layer URL: -
Nimbus will now attempt to connect to a locally running execution client even when the options
--el
and--jwt-secret
are not specified. This is made possible by the following proposed standard:Please note that the standard hasn't been implemented in any execution client yet.
-
Nimbus now support the latest version of the Builder API, adding support for the Capella hard-fork:
-
Improved diagnostic messages and more spec-compliant behavior of the Nimbus validator client when being paired with a non-synced or optimistically synced beacon nodes:
-
The Sqlite3 database engine has been upgraded to version 3.40.1:
-
The doppelganger detection now acts safer after a period of lost network connectivity
-
The doppelganger detection now acts safer in the presence of out-of-order responses from the beacon node:
-
Nimbus can now export ERA files for the Sepolia network:
-
The
--history=prune
mode will no longer interfere with serving light client data for the full retention period as mandated by the spec: -
Nimbus now downloads a longer range of recent execution blocks in order to avoid potential situations where our
Eth1Data
votes fail to agree with the honest majority in the network: -
Nimbus has addressed a potential interruption of deposit syncing when connected to Geth over WebSocket:
Nimbus v23.2.0
is a low-urgency
upgrade providing full support for the upcoming
Capella hard-fork on the Sepolia testnet. Keep an eye out for future mainnet releases!
-
Status now provides an APT repository that will host the latest version of the Nimbus beacon node and validator client software: https://apt.status.im/
-
The
deposits import
command now provides the option--method=single-salt
which will significantly improve the keystore loading speed on start-up on beacon nodes and validator clients running with a very large number of validators. Please see the documentation provided in the Nimbus guide in order to understand the security implications of using the option: https://nimbus.guide/keys.html#optimised-import-for-a-large-number-of-validators status-im#4372 -
More efficient sync committee caching strategies bring 20-30% of syncing speed improvement post Altair: status-im#4592
-
Nimbus performs fewer interactions with the EL node during optimistic syncing which further improves the syncing speed: status-im#4591
-
The Keymanager API now supports all
gas_limit
end-points: https://ethereum.github.io/keymanager-APIs/#/Gas%20Limit status-im#4612 -
Nimbus serves light client updates up to the retention period mandated by the spec even when pruning is enabled: status-im#4499
-
The Linux packages of Nimbus no longer depend on
lsb-release
: status-im#4597 -
The list of bootstrap nodes for the Gnosis network has been expanded: status-im#4603
-
Nimbus now performs fewer
forkchoiceUpdated
Engine API calls with lower risk of reporting conflicting data to the EL node: status-im#4609 status-im#4614 status-im#4623
-
Nimbus will no longer suffer from performance issues when a large number of non-active validators are imported in the beacon node or the validator client: status-im#4590
-
Nimbus will no longer crash when it fails to resolve the hostname of a remote validator imported through the Keymanager API: status-im#4590
-
The Nimbus validator client won't attempt to perform sync committee duties when the attached beacon node is only optimistically synced: status-im#4622
Nimbus v23.1.1
is a high-urgency
hotfix for users who have already enabled block
history pruning after upgrading to v22.3.0
. It fixes an issue where the client may
fail to start after the database has been pruned.
-
A crash on start-up after running the client with
--history:prune
option - "backfill block must have a summary". status-im#4554 -
The
validator-monitor-details
option was accidentally enabled by default in 23.1.0, which lead to significant increase in resource usage.
Nimbus v23.1.0
is a low-urgency
upgrade, introducing support for on-the-fly database pruning making the storage requirements of Nimbus much more predictable on long-term time scales. When pruning is enabled, a typical beacon node expected to consume around 60 to 70 GB of storage. To take advantage of the new functionality without facing any downtime, users are advised to sync a fresh node through trusted node sync which now features a quicker history backfilling implementation.
-
After a trusted node sync, Nimbus requires less time to complete the block backfilling process by downloading the minimum number of historical blocks mandated by the spec:
-
Nimbus is able to sync in optimistic mode with the network even when not paired with an execution layer client. Please note that this mode is not suitable for validating:
-
A new
--history=<archive|prune>
configuration parameter controls the retention of old historic blocks in the database of the client. Enabling pruning on an existing installation will introduce a significant delay on the first run, while history pruning is taking place, so we recommend starting with a fresh database by executing a trusted node sync: -
The validator monitor is now considered out of BETA and enabled by default. To keep the number of created metrics to a reasonable level on installations with large number of validators, the default implies the previous behavior of the
validator-monitor-totals
flag: -
Full support for the latest Capella/Shanghai devnets:
-
Out of date metadata for the Gnosis network bootstrap nodes:
-
Potential hanging of the client caused by inappropriate activation of the TTD block detection on beacon nodes created after the merge:
-
Inappropriate case-sensitivity in the
--log-level
parameter, accidentally introduced in the 22.12.0 release.
Nimbus v22.12.0
is a medium-urgency
release which improves the doppelganger detection in the Nimbus validator through the use of standardized APIs that are compatible with all third-party beacon nodes. Furthermore, it addresses several inconsistencies in the behavior between the stand-alone beacon node and the validator client. This release also allows users of trusted node sync to skip downloading the entire history of validator deposits by syncing against a server that supports the standardized /eth/v1/beacon/deposit_snapshot
REST endpoint.
-
60% more efficient block replaying speed brings faster REST responses and more resilience on the network in the face of heavy forking activity and non-finalization:
-
Support for obtaining a deposit snapshot during trusted node sync from servers supporting the standardized
/eth/v1/beacon/deposit_snapshot
REST endpoint: -
Official docker images for the Nimbus validator client are now available:
https://hub.docker.com/r/statusim/nimbus-validator-client status-im#4439
-
The
skip_randao_verification
query parameter is now also supported in the/eth/v1/validator/blinded_blocks/{slot}
API endpoint: -
The doppelganger detection in the Nimbus validator client is now based on the standardized
/eth/v1/validator/liveness/{epoch}
REST endpoint: -
The validator client will now use with the standard exit code
129
in case of detected doppelganger on the network:
-
A potential false-positive in the doppelganger detection logic:
-
A potential hang in trusted node sync:
-
The built-in support for the Ropsten testnet has been removed:
You can still connect to the Ropsten network by specifying its metadata directory on the command line through the
--network
parameter. -
The
statusim/nimbus-eth2
docker image no longer includes thenimbus_validator_client
binary and the shell scripts included in the release tarballs. Please use the new dedicated image for the Nimbus validator client.
Nimbus v22.11.1
is a high-urgency
hotfix for all users who are
running validators backed by remote keystores and web3signer.
This update addresses a compatibility issue that may result in
missed block proposals. If you are not using a remote signer,
you can safely skip this release.
-
Incompatible encoding used in the web3signer block signing requests after the merge: status-im#4407
-
Ignored
graffiti
option of the validator client: status-im#4417
Nimbus v22.11.0
is a low-urgency
release, bringing the first
production-ready version of the Nimbus validator client.
The validator client will enable advanced users such as professional
institutional operators to take advantage of features such as support
for redundant beacon nodes, enhanced validator privacy and
distributed keystores. Meanwhile, existing users still can use
the streamlined stand-alone beacon node setup, as this mode of operation
will continue to be supported and improved.
-
The Nimbus validator client graduates from BETA to production-ready status after numerous improvements and fixes:
-
The Nimbus beacon node is now compatible with validator clients taking advantage of the
/eth/v1/beacon/blinded_blocks
end-point: -
The validator keystore loading during start-up is now 3x faster:
-
The doppelganger detection now protects validators added on the fly through the Keymanager API:
-
Fine-tuned internal task scheduling can provide better quality-of-service protections for performing all validator duties in the presence of syncing peers performing large number of concurrent block download requests:
-
The
--optimistic
mode of the beacon node allows you to stay synced with the network even without an execution layer node:https://nimbus.guide/optimistic-sync.html#optimistic-mode status-im#4262
Please note that this mode is less secure and intended only for non-critical information retrieval through the Beacon API.
-
The Nimbus build for the Gnosis chain (available when compiling from source) has full support for the upcoming merge:
status-im#4330 https://forum.gnosis.io/t/gip-16-gnosis-chain-xdai-gnosis-merge/1904
-
The stand-alone light client (available when compiling from source) now preserves progress between restarts:
-
A small risk for missing block proposals was present under heavy forking activity in the network due to potential inclusion of inappropriate attestations in the produced blocks:
-
A new batching approach for the registration of validators in the builder network will prevent the creation of requests of excessive size that might be rejected by the network. This was a problem affecting testnet nodes running with thousands of validators:
-
The
--finalized-checkpoint-block
has been deprecated. The client will now download only the latest finalized state during trusted node sync: -
The end-points
/eth/v1/beacon/blocks/{block_id}
,/eth/v1/debug/beacon/states/{state_id}
, and/eth/v1/validator/blocks/{slot}
have been deprecated since they are no longer usable since the Altair hard-fork:
Nimbus v22.10.1
is a low-urgency
point release introducing support for the official light client REST API and improving the stability of Nimbus when paired with an external block builder.
- Support for the official light client REST API: ethereum/beacon-APIs#247 status-im#4213 status-im#4232
-
Nimbus was slowly leaking file descriptors when paired with an external builder: status-im#4235
-
Nimbus could potentially crash under a poor network connectivity to the external builder: status-im#4222
Nimbus v22.10.0
is a medium-urgency
release, continuing our briefly accelerated release schedule and bringing further stability and performance improvements after the merge.
-
Faster block production, bringing practical benefits on low-powered devices such as the Raspberry Pi: status-im#4184 status-im#4196
-
The Nimbus validator client can now work with multiple beacon nodes with configurable responsibilities: status-im#4113 status-im#4140
-
The
/eth/v2/validator/blocks/{slot}
API now features an optionalrandao_reveal
parameter in accordance to the latest Beacon API spec: ethereum/beacon-APIs#222 status-im#3837 -
The
/eth/v1/beacon/blocks
API now supports SSZ-encoded payloads: status-im#4154 -
The new metrics
beacon_block_builder_proposed
,beacon_block_builder_missed_with_fallback
andbeacon_block_builder_missed_without_fallback
can help you track the successful and failed attempts to use the configured external block builder: status-im#4158
-
Rare, but critical conditions manifesting primarily in the Goerli network were leading to an unrecoverable database corruption: status-im#4174 status-im#4192
-
If the chain was re-orged while Nimbus is shut down, this created a low risk that the client may become stuck on a non-canonical block: status-im#4161
-
Nimbus was not serving the best possible light client updates when back-filling after a trusted node sync: status-im#4195
- The pre-altair REST API paths
/eth2/beacon_chain/req/beacon_blocks_by_{range,root}/1/
are now deprecated and will be removed in the next Nimbus version. Since these APIs support only phase0 responses, it is unlikely that there are any remaining clients using them.
Nimbus v22.9.1
is a medium-urgency
upgrade addressing several frequently reported issues after the merge and bringing minor performance improvements in the post-merge world.
- Nimbus no longer supports the non-standard
/api/
prefix for the Beacon REST API. All users should migrate to the standardized/eth/
prefix: status-im#4115
-
Implemented the
/eth/v1/validator/register_validator
, enabling the use of an external block builder when the Nimbus beacon node is used with a validator client: status-im#4115 -
The expensive TTD block detection is no longer performed when the client is launched after the merge: status-im#4152 status-im#4129
-
Peer scoring improvements will result in a more stable peer connectivity during syncing: status-im#3381 status-im#4090
-
The Nimbus peer metrics can now properly track the number of Lodestar peers: status-im#4108
-
Fee recipient configuration, applied through the Keymanager API remains active even after disabling the Keymanager API in a consecutive run: status-im#4078
-
Improved support for working with custom networks: status-im#4132 status-im#4134
-
Using an HTTP connection to the EL client will no longer result in sporadic crashes: status-im#4125
-
Nimbus will no longer trigger warnings or errors regarding an invalid terminal block hash during transition configuration exchanges: status-im#4126
-
The initial transition configuration exchange is performed after a short delay to give more time for the EL client to initialize when all services are started at the same time: status-im#4114
-
The Nimbus beacon node service installed by our DEB and RPM packages will now use the correct Engine API port by default (8551 instead of 8546): status-im#4099
-
Nimbus has better compatibility now with various beacon API servers used for trusted node sync (such as Prysm and Alchemy): status-im#4133 status-im#4139
-
Nimbus was delivering incorrect head block details through the events API: status-im#4119 status-im#4141
-
Nimbus can now import keystores exported from ethdo or Prysm: status-im#4149
Nimbus v22.9.0
is a high-urgency
upgrade that fixes a critical pre-TTD block production issue affecting users that restarted their node after Bellatrix. It also improves compatibility with Besu, Prysm and slow block builders and provides a speed boost in block processing important for those running on Raspberry Pi and similar hardware.
With the merge drawing near, the focus of this release has been to include low risk changes that improve stability and compatibility - if you are unsure whether to upgrade, do reach out to us in discord to discuss your particular deployment.
A shout out to our great community for reporting and helping diagnose the issues that led up to this release - in particular Michael Sproul (Lighthouse) and Joe Clapis (Rocket Pool).
-
Allow more time for block builder to deliver block #4088
-
Improve Bellatrix block processing performance #4085 and #4082
-
Optimize execution layer calls when not producing blocks, improving Besu performance and compatiblity #4055
-
Revise timing of execution layer configuration call, resolving warnings that no consensus client is present on Geth and Besu #4077
-
Log
Exchanged engine configuration
when first connected to correctly configured execution engine #4096 -
Switch to
nim-websock
for websocket connections, resolving delays when payloads exceed 1mb #4061
-
Fix pre-TTD block proposals on nodes that (re-)started after Bellatrix #4094
-
Fix gossip message id, improving connectivity health with Prysm #4076
-
Improve handling of blocks deemed invalid by the execution layer #4081
-
Fix a rare crash that could happen when execution layer disconnected #4095
Nimbus v22.8.2
is a low-urgency
hotfix release, eliminating a risk for potential crash during block production that was introduced in the v22.8.1
release. You can safely skip this release if you haven't enabled DEBUG logging on your beacon node, as the risk exists only when DEBUG logging is enabled.
- Reduced CPU usage for validator registration when using an external builder: status-im#4040
- A potential crash during block production when DEBUG logging is enabled: status-im#4054
Nimbus v22.8.1
is a high-urgency
upgrade, improving the stability and performance of Nimbus in post-merge networks. Upgrading is highly recommended due to improved timing of the interactions with the execution engine which may lead to higher profitability from block production, especially for users running Nethermind.
-
More timely block proposals in the presence of a non-responsive builder node: status-im#4012
-
More timely delivery of fork-choice update information to the execution client, enabling the production of more profitable blocks: status-im#4012
-
Improved SHA256 hashing performance resulting in a minor overall CPU usage reduction: status-im#4017
-
Reduced latency in the light client when following the head of the chain optimistically: status-im#4002
-
Spec-compliant delivery of the "safe block hash" property of the "fork-choice update" messages sent to the Engine API: status-im#4010
-
Relax overly aggressive gossip filtering conditions for incoming blocks: status-im#4044
-
New metrics
beacon_block_production_errors
andbeacon_block_payload_errors
for detecting non-healthy operation of the Engine API: status-im#4036
-
Sporadic loss of connectivity to the execution engine in the presence of large payloads: status-im#4028
-
Inappropriate loss of connectivity to honest peers in the presence of a non-responding execution client: status-im#4020
-
A loophole allowing the inclusion of very old and invalid slashing and exit messages within blocks: status-im#4013
-
Confusing error message when trusted node sync is executed with an invalid REST URL: status-im#4024
Nimbus v22.8.0
is a medium
urgency release, featuring full support for the upcoming mainnet merge! All users should upgrade at their earliest convenient, but no later than 5th of September.
Since the network will go through the Bellatrix hard-fork on Sept 6, 2022, 11:34:47am UTC, failure to upgrade in time will result in inactivity penalties.
Please note that once the network reaches the terminal total difficulty (currently estimated to happen between 13th and 15th of September), it will no longer be possible to operate a beacon node without pairing it with a single non-shared merge-ready execution client. Nimbus is fully compatible will all execution clients and the required configuration steps for all of them are the same. Please refer to our merge guide for more details:
https://nimbus.guide/merge.html
To raise awareness of the required configuration changes, once the Bellatrix fork is activated on 6th of September, Nimbus will refuse to start unless a properly configured and authenticated Engine API end-point is provided through the command-line options --web3-url
and --jwt-secret
. If you need more time to complete the transition, you can temporarily run the beacon node with the command-line option --require-engine-api-in-bellatrix=no
, but please note that such a setup will stop working once the network TTD is reached!
We would like to say a huge THANK YOU to all of our users who provided immensely valuable feedback in the many months of hard work leading to the merge and to all the fellow research and implementation teams who made this historic release possible!
Onwards and happy merging!
-
Nimbus will refuse to start unless connected to a properly configured execution client in Bellatrix-enabled networks: status-im#4006
-
The custom error code returned by Nimbus when a validator doppelganger is detected has been changed from 1031 to 129 to improve compatibility with
systemd
: status-im#3977
-
Support for external block builders (a.k.a. MEV): status-im#3883
-
Beta release for the Nimbus stand-alone light client, which can be used to drive any execution client without requiring a full-blown beacon node: https://nimbus.guide/light-client-data.html
-
The first spec-compliant implementation of the LibP2P protocols for serving light client data: https://nimbus.guide/light-client-data.html
-
Keystore locking prevents accidentally loading the same validator keys in multiple instances of the Nimbus beacon node and the Nimbus validator client, thus eliminating a potential slashing risk: status-im#3907
-
Debian and RPM packages for the Nimbus beacon node and the Nimbus validator client are now available as part of the release. In the near future, Status will also provide a package repository, offering a more convenient installation: status-im#3974 status-im/infra-nimbus#79
-
Improved performance on networks with heavy forking activity through a reduction of the required state replays: status-im#3990
-
The Nimbus validator client now supports validator activity metrics such as
beacon_attestations_sent
,beacon_aggregates_sent
,beacon_attestation_sent_delay
,beacon_blocks_sent
,beacon_blocks_sent_delay
,beacon_sync_committee_messages_sent
,beacon_sync_committee_message_sent_delay
,beacon_sync_committee_contributions_sent
: status-im#3915 -
The sync status displayed in the Nimbus status bar and certain log messages now describes the state of the client more accurately (optimistically synced vs fully synced): status-im#3987
-
Spec violation in the expected payload of the
/eth/v1/validator/prepare_beacon_proposer
Beacon API end-point: status-im#3938 -
Invalid empty execution payloads being produced when the execution client is not responding: status-im#3991
-
Potentially incorrect Eth1 block votes, disagreeing with the forming majority: status-im#3944
-
More resilient deposit synchronization when Nimbus is paired with a highly loaded execution client: status-im#3943 status-im#3957
-
A potential delay in detecting the terminal total difficulty block: status-im#3956
-
Missing Gossip filtering rule for sync committee contributions resulting in unnecessary traffic: status-im#3941
-
Compatibility issue preventing trusted node sync from Lodestar nodes: status-im#3934
-
A potential crash while processing rare gossip messages such as slashings and exits: status-im#3965
-
Inappropriate attestations sent by the validator client when the connected beacon node is only optimistically synced: status-im#3968
Nimbus v22.7.0
is a low
urgency release packing everything necessary for the upcoming Prater/Goerli merge and introducing the Nimbus validator client (currently in BETA).
-
Support for fee recipient management through the Keymanager API and through the
/eth/v1/validator/prepare_beacon_proposer
Beacon API end-point: status-im#3864 status-im#3901 -
Support for the post-merge optimistic sync specification: status-im#3793
-
More comprehensive spec-compliance in our fork-choice implementation: status-im#3849
-
More spec-compliant handling of
QUANTITY
values within the Engine API responses: status-im/nim-web3#55 status-im#3844 -
The
Slot end
log message now includes information regarding current and upcoming sync committee duties to help you identify the most appropriate time to restart the client during an upgrade: status-im#3854 -
Specifying a
WEB3_URL
environment variable is no longer mandatory when launching beacon nodes with therun-*-beacon-node.sh
scripts: status-im#3810 -
The
--finalized-checkpoint-state
and the--finalized-checkpoint-block
command-line parameters can no longer be used with certain invalid inputs: status-im#3858 -
Specifying
--network=goerli
is now equivalent to specifying--network=prater
: status-im#3874
-
A risk for invalid block proposals during high forking activity in the network due to inappropriate inclusion of attestations from other forks: status-im#3893
-
Interrupted tracking of deposits, triggered by a non-responsive web3 end-point: status-im#3905
-
Inappropriate error returned by the REST API when broadcasting of Gossip messages is not immediately possible: status-im#3843
-
Rare conditions under which P2P connections were closed inappropriately: status-im#3795
-
Potential inaccuracies in the
next_action_wait
metric: status-im#3862
Nimbus v22.6.1
is a low-urgency
release which comes pre-configured with the correct TTD value for the Sepolia network and improves the behavior of Nimbus in merge testnets.
- Allow testing the Engine API JWT credentials even before the merge: status-im#3786
-
Lack of detection of the connected execution client's network when attached to the Engine API port: status-im#3804
-
Logic error leading to a premature start of the
exchange transition configuration
Engine API requests: status-im#3809 -
Inappropriate inclusion of the
execution_optimistic
field in REST responses before the merge: status-im#3807
Nimbus v22.6.0
brings support for the merge testnets Ropsten and Sepolia (please stay tuned for TTD announcements for the latter) and a lot of polish where we've taken the time to address a long list of UX improvements and bug fixes suggested or reported by our users. We are deeply grateful to everybody who contributed valuable feedback for this release.
-
TTD detection and Panda art for the merge! status-im#3670 status-im#3745
-
The execution layer priority fees recipient address can be configured individually for each validator: status-im#3652
-
Through better defaults, the parameters
--rest-url
,--trusted-node-url
can be omitted if the targeted node is running on the same machine: status-im#3689 -
Improved spec-compliance with the Beacon API and the Engine API as defined after the merge: status-im#3679 status-im#3780
-
The custom error code
129
will signal a detected doppelganger on the network. This can be handled in the Nimbus's service supervisor to prevent an automatic restart: status-im#3728 -
The Nimbus status bar can be configured to display the current version number: status-im#3747
-
Specifying the
--terminal-total-difficulty-override
parameter is no longer necessary for the Ropsten network: status-im#3754 -
Built-in support for the Sepolia network which will launch on June 20th and reach TTD shortly after: status-im#3762
-
More robust syncing with the connected execution layer node in Bellatrix-enabled networks: status-im#3759
-
The
web3 test
command is now compatible with nodes that have been configured to serve only the Engine API: status-im#3761
-
A rare crash triggered when using a HTTP web3 URL: status-im#3669
-
ERA checkpoint sync failing with "Backfill block must have a summary": status-im#3675
-
Incorrect sync progress indicator shortly after a trusted node sync: status-im#3736
-
Incorrect values returned by the
/eth/v1/node/syncing
API under rare circumstances: status-im#3720 -
Misleading log message when an attestation was not delivered to any peer: status-im#3737
-
Incorrect handling of case-sensitive web3 URLs: status-im#3757
-
Incorrect encoding of the
current_epoch_participation
andprevious_epoch_participation
fields in the REST requests returningBeaconState
results: status-im#3776 -
Incorrect URL for the Keymanager delete keystores request: status-im#3727
-
Non-standard encoding required by the Keymanager API for the import keystores request: status-im#3768
-
A significant source of omitted events in the REST events API: status-im#3664
-
Incorrect parsing of the
weak-subjectivity-checkpoint
parameter: status-im#3765 -
Lack of support for trailing commas in lists and inline tables in the TOML config files: status-im/nim-toml-serialization#47
- The Nimbus-specific JSON-RPC service which was deprecated in version v22.3.0 is now removed. If you are currently relying on the JSON-RPC API, please consider switching to the official REST API. Using any of the
--rpc
flags will now result in a warning: status-im#3656
Nimbus v22.5.2
is a low-urgency
maintenance release updating Ropsten testnet support.
-
Modify proposer boost from 70% to 40% to improve network consensus: https://github.com/status-im/nimbus-eth2/commit/14dc3855f6cd06579294322a6ed206f678c8530f
-
Update Ropsten TTD to a large enough number it can't be readily triggered by mining: status-im#3668
Nimbus v22.5.1
is a low-urgency
maintenance release addressing a Web3 compatibility regression and introducing Ropsten testnet support.
- Support for the Ropsten testnet, intended for merge testing: status-im#3648
-
Restore compatibility with certain Web3 endpoints: status-im#3645
-
More spec-compliant handling of JSON fields in REST, for better compatibility with added and optional fields: status-im#3647
Nimbus v22.5.0
is a low-urgency
maintenance release. It implements the proposer boosting fork-choice policy and is compliant with the latest Bellatrix specifications. It also provides an early preview of our built-in support for BLS threshold signatures (via regular Web3Signer instances): this marks the first step of our long-term secret-shared validators roadmap which enables node operators / staking pools to deploy Nimbus in secure high availability setups (guaranteeing ~100% uptime).
-
A safer fork-choice algorithm which implements the proposer boosting policy: ethereum/consensus-specs#2353 status-im#3565
-
A completely revamped snappy implementation which brings significant speed-ups: status-im#3564
-
Support for the latest Bellatrix specifications (a.k.a. The Merge) + all Kiln testnets: status-im#3590
-
An initial preview release fеaturing built-in support for distributed keystores, (part of our secret shared validators roadmap): status-im#3616
-
Reduced CPU usage when serving blocks to other syncing clients: status-im#3598
-
A more spec-compliant implementation of the
/eth/v1/config/spec
REST end-point (implementing the v1.1.10 version of the spec): status-im#3614 -
Improved compatibility with all versions of Web3Signer: status-im#3640
-
The potential for missed block proposals in the case where an invalid deposit is submitted to the deposit contract: status-im#3607 status-im#3639
-
A crash triggered by the use of Web3Signer remote keystores: status-im#3616
-
A rare crash triggered when Nimbus is performing a large number of concurrent HTTP requests: status-im/nim-chronos#272 status-im/nim-chronos#273
Nimbus v22.4.0
is a low-urgency
upgrade which brings with it further optimisations, and better user experience around trusted node sync. It lays the foundations for upcoming the merge hard-fork which will be fully supported in our next release (v22.5.0
).
-
All CPU cores are now used by default: previously enabled by passing
--num-threads:0
on the command-line: status-im#3493 -
250 MB reduction in memory usage: thanks to more efficient data structures for the finalized portion of the chain history: status-im#3513
-
Higher performance historic queries (using REST API) after trusted node sync: Nimbus now re-indexes the backfilled chain of blocks: status-im#3452
-
Broadcasted attestations are more likely to be included in blocks by other nodes: thanks to a tweak to the attestation sending time: status-im#3518
-
The REST API now only returns current and relevant information in response to VC queries: in other words, information from the recent non-finalized portion of the chain history: status-im#3538
-
Better and more consistent gossip mesh health: the
--max-peers
option now works as a target that can be exceeded by the client temporarily in order to maintain good gossip mesh health; the newly introduced--hard-max-peers
option now acts as the hard limit that should not be exceeded (default set tomax-peers * 1.5
): status-im#3346 -
An ERA files developer preview: ERA files are an ultra-efficient long-term storage format for finalized chain history: https://github.com/status-im/nimbus-eth2/blob/unstable/docs/e2store.md
-
Nimbus no longer crashes when a HTTP URL is specified as a
--web3-url
end-point: status-im#3582 -
The REST end-point
/eth/v1/beacon/headers
is now able to return backfilled blocks: status-im#3472 -
The Nimbus status bar has been disabled on Windows in order to avoid sporadic hangs in certain terminal emulators: status-im#3484
-
A large start-up delay after backfilling: status-im#3516
-
A rare problem which prevented the node from starting successfully after a trusted node sync: status-im#3517
-
Confusing error messages when Nimbus lacks the necessary file system permissions to create its database: status-im#3536
- The support for the Pyrmont testnet has been removed in order to reduce the Nimbus binary size: status-im#3568
Nimbus v22.3.0
is a low-urgency
upgrade that marks the beginning of a more predictable release cadence for Nimbus. Going forward, we'll be publishing a new release each month, following a feature freeze period with intensified testing and monitoring of the introduced code changes on our dispersed fleet of mainnet validators.
Please note that the new versioning scheme is tied to the calendar. The number 22 indicates the year of the release (2022), while 3 is the month (March). The last digit is the patch number of the release and it will have a non-zero value only when we ship a hotfix during the month.
-
Nimbus can now run as a service on Windows: use the
--run-as-service
flag: status-im#3441 -
All command-line options can now be provided in a configuration file: use the
--config-file
flag: status-im#3442 https://nimbus.guide/options.html -
Lower CPU and bandwidth usage, thanks to better handling of already-seen attestation aggregates: status-im#3439
-
Reduced memory usage for nodes bootstrapped with trusted node sync: status-im#3429
-
Reduced performance on Windows due to the use of a less efficient method for collecting stack traces status-im#3466
-
Non-spec-compliant URLs in the Keymanager APIs for handling remote keystores https://github.com/status-im/nimbus-eth2/commit/4c01b777736f0d5d6fe38b37a4349741f6944e4c
-
Extremely slow slashing DB import for validators with long validation history: the import should be nearly instant now status-im#3393
-
Validator index-out-of-bounds crash that was triggered upon certain requests to the
/eth/v1/beacon/states/{state_id}/validators/{validator_id}
API status-im#3463 -
An off-by-one logic error preventing sync committee messages to be published in the first slot of each sync committee period https://github.com/status-im/nimbus-eth2/pull/3470/commits/542e645bedec7702a973dc5cdaae87175e353009
- The JSON-RPC service (
--rpc
flag) option is now deprecated. It's scheduled for removal in versionv22.6
(i.e. June of this year). If you are currently relying on the JSON-RPC API, please consider switching to the official REST API.
Nimbus v1.7.0
is a low-urgency
feature-packed upgrade, which brings support for trusted node sync (also known as checkpoint sync) and HTTPS web3 providers.
Of particular note: the Keymanager API now supports remote keystores (a.k.a web3signer keystores).
-
Nimbus will no longer rewrite HTTP(S) web3 URLs to their respective WebSocket alternatives. Please review your setup to ensure you are using the desired web3 end-point.
-
The peer scoring has been further tuned. As such the
--max-peers
should not be set below 70. Note that Loweringmax-peers
does not significantly improve bandwidth usage, but does increase the risk of missed attestations.
-
Full support for HTTP and HTTPS web3 URLs: status-im#3354
-
Nimbus now treats the first
--web3-url
as a primary and preferred web3 provider. Any extra URLs are treated as fallback providers (to be used only when the primary is offline). As soon as the primary is usable again, Nimbus will switch back to it. -
The Keymanager API now supports management of remote keystores (also known as web3signer keystores): status-im#3360 * The typical memory usage of Nimbus on mainnet is now below 1GB: status-im#3293
-
128MB of savings come from exploiting a provision in the official spec, which allows clients to respond with only non-finalized blocks to network queries which request blocks by their root hash.
-
Faster beacon node startup-times: status-im#3320
-
The REST API is now compatible with CORS-enabled clients (e.g. browsers): status-im#3378
-
Use the
--rest-allow-origin
and/or--keymanager-allow-origin
parameters to specify the allowed origin. -
A new
--rest-url
parameter for thedeposits exit
command: status-im#3344, status-im#3318 -
You can now issue exits uing any beacon node which provides the official REST API. The Nimbus-specific JSON-RPC API will be deprecated in our next release, with a view to completely phasing it out over the next few months.
-
The REST API will now returns JSON data by default which simplifies testing the API with
curl
. -
The notable exception here is when the client requests SSZ data by supplying an
Accept: application/octet-stream
header. -
Fairer request capping strategy for block sync requests and reduced CPU usage when serving them: status-im#3358
-
More accurate Nim GC memory usage metrics.
-
BLST upgrade (latest version): status-im#3364
-
The
web3 test
command now provides more data about the selected provided: status-im#3354
- Unnecessary CPU and bandwidth usage: status-im#3308
- The result of staying subsribed to sync committee topics even when there were no validators in the committee.
- Excessive logging on beacon nodes with large numbers of validators (in particular, those with
--validator-monitor-totals
enabled): status-im#3332 - Deviations from the spec in the REST API; this led to sub-optimal performance when Nimbus was paired with Vouch.
- Naming inconsistencies in the "totals" metrics (this was produced by the validator monitor).
- Non-compliant implementation of the
/eth/v1/node/health
API (we were not producing HTTP status codes as mandated by the spec). - Unnecessary restarts of the Eth1 syncing progress when the web3 provider connection was lost during sync: status-im#3354
Nimbus v1.6.0
is a low-urgency
performance optimisation release with improved peer management.
v1.6.0
adds support for the Keymanager
API (currently in BETA):
https://nimbus.guide/keymanager-api.html
As well as a comprehensive set of metrics for validator performance monitoring:
https://nimbus.guide/validator-monitor.html
- Tuned peer management: reduces the likelihood of missed attestations
- If you've seen frequent "No peers for topic" in your logs, this release will help
- Improved buffer management in the networking layer: reduces both CPU and memory usage.
- Further optimised batch verification of gossip messages: provides a 2-fold improvement in throughput.
- Comprehensive set of metrics for live validator performance monitoring in Grafana and support for producing detailed historic reward analysis in
ncli_db
(note thatncli_db
is available only when compiling from source at the moment). - Support for the new Keymanager API: add, remove, and migrate validators on the fly (BETA).
- Blazingly fast historical traversals in the REST API for beacon chain data mining: state caching brings up to a 10x speed-up in some common usage patterns (e.g. obtaining historic data slot by slot or epoch by epoch).
- 3x speed-up in snappy compression and decompression.
- Support for obtaining JSON payloads from the REST API.
Nimbus v1.5.5
is a medium-urgency
bugfix release which contains a number of significant optimisations; of particular note is a 6x speed-up in epoch processing and 2x speed up in Altair block processing.
In addition, v1.5.5
adds support for the web3signer
protocol (currently in BETA).
-
The potential for missed block proposals when a third-party validator client (with at least one imported validator) is used with a Nimbus beacon node (with no imported validators)
- The web3 connection not being enabled when running third-party validator clients
-
A rare condition in which the REST service becomes unavailable.
-
Inappropriate error messages produced by the REST API: when a validator client is publishing the same attestations or sync committee messages through multiple beacon nodes.
- 6x speed-up in epoch processing: status-im#3089
- 2x speed up in Altair block processing: status-im#3115
- A 12% (minimum) reduction in the outgoing GossipSub traffic: status-im#3112
- Across the board performance improvements in the REST API: status-im#3092
- The REST API can now report sync committee information for the next sync period: status-im#3133
- Added support for the web3signer protocol (beta release): status-im#3077
Nimbus v1.5.4
is a medium-urgency
hotfix release. It addresses an important issue which, in rare cases, can lead to the loss of attestations and sync committee messages. This can, in turn, lead to a reduction in rewards.
Please upgrade at your earliest convenience.
-
A rare issue during the construction of sync committee contributions: invalid BLS aggregate signatures were being produced under certain conditions; this had the potential to negatively affect the peer score of the node, and impact its ability to deliver gossip messages.
-
A non-spec-compliant implementation of the
/eth/v1/validator/duties/sync/{epoch}
REST API. -
A crash in the
/eth/v2/debug/beacon/states
REST API call on systems with limited stack space.
-
A nice little performance improvement for block verification and replay.
-
Improved error messages in the REST API.
-
The
/eth/v1/config/spec
REST API now returns more information regarding spec config parameters.
-
The
--log-file
option is now deprecated and may be removed in a future release (if you wish to log to a file, we recommend redirecting the standard output).Please note that the --log-file option was previously supported only when Nimbus was built from source. If your existing configuration used the --log-file option with a binary release, upgrading to v1.5.4 will enable the log file creation (though a deprecation warning will be printed on start-up).
Nimbus v1.5.2
is a high-urgency
release for all users who haven't yet upgraded to the v1.5.x
series
which add support for the upcoming beacon chain Altair hard-fork.
Please upgrade as soon as possible - before the 27th of October.
For users already running v1.5.1
, the release brings a number of important bug fixes and optimizations
and upgrading is still highly recommended.
-
Faster ramp up of peers when starting the beacon node.
-
Added new metrics for keeping track of dropped gossip messages:
https://github.com/status-im/nimbus-eth2/commit/bf6ad41d7dfd0899527a0374009a3fcf2a32361b
-
The run-*-node.sh scripts provided in out Github repository will now enable the --rest and --metrics options by default.
-
Potential crashes triggered by certain JSON-RPC and REST API requests.
-
Unnecessary source of syncing time when the beacon node is restarted after the Altair transition.
-
Certain non spec-compliant responses of the REST API.
Nimbus v1.5.1
is a high-urgency
release for all users who have already upgraded to v1.5.0
. It fixes a deposit contract syncing issue manifesting as the warning message "Eth1 chain not synced". Under specific circumstances, such a failure to sync the Eth1 chain may result in missed Eth2 block proposals, so affected users are advised to upgrade as soon as possible.
For anyone still running Nimbus v1.4.x or earlier, migrating to v1.5.1 continues to be a medium-urgency
but mandatory upgrade that must be installed before October 27th.
Nimbus v1.5.0
is a medium-urgency
but mandatory upgrade which adds support for the upcoming beacon chain Altair hard-fork.
Please upgrade at your earliest convenience - before the 27th of October.
N.B You must upgrade before October 27th in order to follow the mainnet beacon chain. Failure to do so will result in downtime penalties.
This is the first hard fork for the beacon chain, and while a significant amount of testing has been conducted in the run up to this release, we recommend users regularly monitor our announcement channels (discord and/or the newsletter) for potential updates.
To celebrate the feature complete release candidate of our REST API, we've opened up the ports on some of our fleet nodes - don't do this at home ;)
In order to interact with these, you should point your apps and tools to:
- http://unstable.mainnet.beacon-api.nimbus.team/ -
mainnet
data, the latestunstable
branch - http://unstable.prater.beacon-api.nimbus.team/ -
prater
testnet data, the latestunstable
branch
Note that right now these are very much unstable testing instances. They may be unresponsive at times - so please do not rely on them for validation. We may also disable them at any time.
-
Support for the Altair hard fork and the latest Ethereum consensus spec (
v1.1.1
) -
Our REST API is now feature complete, supporting the full Beacon API specification. This API should not be exposed to the public Internet as it includes multiple endpoints which could open your node to denial-of-service (DoS) attacks.
- Known limitations: To use the REST API with a validator client, you need to enable the
--subscribe-all-subnets
option. This requirement will be removed in future versions.
- Known limitations: To use the REST API with a validator client, you need to enable the
-
A new experimental
--num-threads=X
option allows Nimbus to take advantage of multiple CPU cores when verifying attestations. Set it to1
to use one worker thread,2
for two worker threads, etc. The default value is set to1
in this release, but future versions will set it to0
(this tells the client to use as many worker threads as there are CPU cores available). N.B .enabling anything other than1
is considered experimental at this stage. -
Improved peer discovery logic that ensures good connectivity to all gossip subnets.
-
A new
version
metric that makes it easy to keep track of client upgrades within Grafana dashboards. -
New metrics
libp2p_peers_identity
,libp2p_peers_traffic_read_total
,libp2p_peers_traffic_write_total
, subdivided through Prometheus labels by the client type of the peer (i.e. nimbus, prysm, teku, lodestar or lighthouse). -
BLST has been upgraded to the latest version (
v0.3.5
). -
The
--network
option now accepts a directory with custom network metadata in the format of the eth2-testnets repository. TheSECONDS_PER_SLOT
configuration value remains the only setting that must be supplied at compile-time through the-d:SECONDS_PER_SLOT=X
nim compilation flag.
Nimbus v1.4.2
- "Upgrade procedure: Hotfix release"
This release is marked as low-urgency
for all Nimbus users other than those who have recently updated to v1.4.1
from a version earlier than v1.1.0
- for these users this is a high-urgency
release.
This release fixes an issue in the upgrade procedure when upgrading from a version earlier than 1.1.0
to 1.4.x
.
How can I tell if I've been affected?
If you've already upgraded to 1.4.1
, you can tell that you've been affected if you're seeing the following WRN log
:
Received invalid sequence of blocks
To re-iterate, this issue only affects users who are upgrading from 1.0.12
or earlier (released on 2021-03-10
), and have not run any release in between. Everyone else can ignore this release.
Nimbus v1.4.1 - "Every attestation counts"
This release is marked as low-urgency
Nimbus v1.4.0
users might have noticed that they are missing a small number of (seemingly random) attestations since the update. Our investigation into the matter has showed that, due to v1.4.0
's significant performance improvements, Nimbus validators occasionally send their first attestation for a new epoch before some peers are ready. These "slow" peers end up dropping early attestations because they're busy with the epoch transition.
It's a rare occurrence, since it requires a validator to be scheduled to attest in the first slot of an epoch and for the beacon node to only be connected to "slow" peers for the respective libp2p topic. If both these conditions are true, a premature attestation may be lost in time, like tears in the rain.
As a fix, we are using a larger send delay: #2705.
Fo those Nimbus v1.4.0
users who are concerned about reaching optimal attestation effectiveness, we encourage you to upgrade as soon as possible.
Other changes include log flushing and metrics fixes.
Full list:
- increase attestation wait time (#2705)
- ensure logs are printed without delays (#2669)
- fix metrics on Windows (#2707)
This release is marked as low-urgency - please update at your convenience.
It contains improvements to attestation effectiveness and CPU usage.
It also contains improvements to the RPC APIs, as suggested by DappNode and RocketPool, and in preparation for our refactored validator client.
A reminder that if you're running Nimbus with a version prior to 1.0.10
(March 2021) you are exposed to a vulnerability in our core cryptography library (this library is used by all eth2 clients). See this blst security advisory for more information. If this concerns you, please update as soon as you can.
- Nightly builds for the very adventurous. (status-im#2640)
- We expect users of nightly builds to be comfortable providing debugging logs.
- RPC API endpoints (status-im#2585, status-im#2586)
/eth/v1/beacon/pool/attestations
/api/eth/v1/validator/aggregate_and_proofs
- Doppelganger detection: fixed false positive on fast restart (status-im#2656)
- Database read performance improvements during epoch transitions and startup (status-im#2639, status-im#2617)
- Better usage of caches, specially when validating attestations (status-im#2631)
This release offers safer and easier options to migrate to Nimbus from other clients. It also brings further performance optimizations.
We've added:
-
A new
slashingdb
sub-command withimport
andexport
options. This allows for safely migrating to Nimbus from another client (as per the EIP-3076 slashing protection interchange format). Please see the the newly prepared migration guides for the details. -
A new
ncli_db validatorPerf
command. This can be used to perform a textual report for the attestation performance of a particular validator (please note thatncli_db
is available only when compiling from source). -
Official binaries for macOS (AMD64 and ARM64).
-
Pruning of the slashing protection database and a transition to more optimal queries. This results in a significant reduction in both disk and CPU usage on nodes running a large number of validators.
-
More consistent level of validation for the attestations received from third-party sources and the JSON-RPC and REST APIs. This prevents invalid attestations from being broadcasted to the network.
-
Performance tuning of attestation subnet transition timings and state snapshotting intervals. This results in improved CPU and bandwidth usage.
We've fixed:
- Problems in the GossipSub subnet walking logic leading to unnecessary bandwidth and CPU costs.
This is a bugfix release improving the stability of the REST API and addressing issues discovered during the mainnet deposit processing accident of 24-25 April.
New features:
- More efficient attestation processing pipeline using less queuing.
We've fixed:
-
Insufficient validation of third-party Eth1Data votes.
-
Sporadic REST API connection interruptions resulting from large request or result payloads.
-
Incorrectly sent empty GossipSub IWANT messages.
This is a hotfix release that solves the database migration issue highlighted in the previous release -- this problem affected new Nimbus users who used v1.1.0 to sync with the network from genesis, essentially resetting their state database and causing them to start re-syncing from genesis.
If you have used an older version of Nimbus prior to upgrading to v1.1.0, you should not be affected.
If you were affected, you have a couple of options available to you:
-
If you have backed-up your database prior to upgrading to v1.2.0, you can restore the database from backup and execute the migration successfully after upgrading to this release.
-
If you haven't backed up your database, you can upgrade to this release at your convenience; rest assured it won't delete your sync history.
Please accept our sincerest apologies for any inconvenience we may have caused. We are reviewing our release testing policies to ensure that we cover a greater number of possible upgrade paths going forward.
If v1.1.0
was the big I/O update, v1.2.0
is all about the CPU - together, these
updates help secure Nimbus against future network growth, and provide us
with a higher security margin and substantial [profitability improvements]
(https://twitter.com/ethnimbus/status/1384071918723092486).
To highlight just one data point, CPU usage has been cut by up to 50% over v1.1.0 ( 🙏 batched attestation processing). This makes it the first release we can officially recommend for validating on a Raspberry Pi 4.
N.B. this release contains a critical stability fix so please make sure you upgrade!
New features:
-
Beta support for the official Beacon Node REST API: https://ethereum.github.io/eth2.0-APIs/. Enable it by launching the client with the
--rest:on
command-line flag -
Batched attestation verification and other reforms -> massive reduction in overall CPU usage.
-
Improved attestation aggregation logic -> denser aggregations which in turn improve the overall health of the network and improve block production.
-
More efficient LibP2P connection handling code -> reduction in overall memory usage.
We've fixed:
-
A critical stability issue in attestation processing.
-
scripts/run-*-node.sh
no longer prompts for a web3 provider URL when the--web3-url
command-line option has already been specified.
This release brings planned reforms to our database schema that provide substantial performance improvements and pave the way for an an improved doppelganger detection ready immediately to propose and attest to blocks (in a future release).
Please be aware that we will remain committed to maintaining backwards compatibility between releases, but this release does not support downgrading back to any previous 1.0.x release.
As a safety precaution, we advise you to please backup your Nimbus database before upgrading if possible.
New features:
-
More efficient state storage format ==> reduced I/O load and lower storage requirements.
-
More efficient in-memory cache for non-finalized states ==> significant reduction in memory usage.
-
More efficient slashing database schema ==> scales better to a larger number of validators.
-
The metrics support is now compiled by default thanks to a new and more secure HTTP back-end.
-
Command-line tools for generating testnet keystores and JSON deposit files suitable for use with the official network launchpads.
-
setGraffiti
JSON-RPC call for modifying the graffiti bytes of the client at run-time. -
next_action_wait
metric indicating the time until the next scheduled attestation or block proposal. -
More convenient command-line help messages providing information regarding the default values of all parameters.
-
--direct-peer
gives you the ability to specify gossip nodes to automatically connect to. -
Official docker images for ARM and ARM64.
-
Support for fallback
--web3-url
providers.
We've fixed:
-
Long processing delays induced by database pruning.
-
File descriptor leaks (which manifested after failures of the selected web3 provider).
-
The validator APIs now return precise actual balances instead of rounded effective balances.
-
A connection tracking problem which produced failed outgoing connection attempts.
Breaking changes:
-
Nimbus-specific JSON-RPCs intended for debug purposes now have the
debug_
prefix:getGossipSubPeers
is nowdebug_getGossipSubPeers
getChronosFutures
is nowdebug_getChronosFutures
This is bugfix release correcting an error in the Prater testnet config leading to incorrect Eth1 voting.
This is a minor release adding support for connecting to the Prater testnet.
This release contains important security and performance improvements.
Upgraded:
-
We're now running version 0.3.3 of the BLST library: https://github.com/supranational/blst/releases/tag/v0.3.3
-
We've switched to a more recent version of BearSSL (this version features a more up-to-date list of trusted root certificates)
-
We're now consistent with the v1.0.1 Eth2 spec
We've fixed:
-
A frequent crash occurring on certain hardware configurations after building Nimbus from source.
-
Long processing delays triggered by the reception of attestations that reference already pruned states.
-
LibP2P peer management issue which led to an accumulation of inactive connections.
-
A false-positive in doppelganger detection triggered by rebroadcasted older attestations arriving with a significant delay.
New features:
-
A new improved format of the slashing protection database:
-
Significantly reduces the disk load with a large number of validators (1000+).
-
Makes it possible to enhance our doppelganger detection in the future such that waiting for 2 epochs before attesting is not necessary.
To ensure smooth upgrade and emergency rollback between older and future Nimbus versions, v1.0.10 will keep track of your attestation in both the old and the new format. The extra load should be negligible for home stakers.
-
This version was an internal release candidate build for the 1.0.10 release.
This release includes important JSON-RPC stability improvements and compatibility fixes, which make it possible to use Nimbus as a RocketPool operator.
New features:
-
RocketPool integration: see rocket-pool/smartnode#89 and https://github.com/rocket-pool/smartnode-install/pull/26/commits/da720acc8f4c1c31c05971748fbc144de1621830
-
Next attestation time displayed on every "Slot end" log message (helps you select the best time for restarting the node)
-
libp2p scoring: disconnect from badly performing peers and prioritise peers with better latency and throughput.
We've fixed:
-
A rare crash triggered when connecting to a web3 provider using a secure web socket.
-
JSON-RPC spec violations and potential DoS attack vectors.
-
Two stale bootstrap node addresses.
A release which provides additional protection against accidental slashings and further performance improvements across the board.
New features:
-
New slashing protection mechanism (doppelganger detection) prevents your validator from contradicting itself if you have accidentally left it running on another machine (see the
--doppelganger-detection
option). -
Optimized batching of BLS signature verification leading to faster sync speeds and reduced CPU load.
-
Further improvements to attestation subnet walking resulting in a reduction in both bandwidth and CPU usage.
-
A new
--subscribe-all-subnets
option allowing the node to maintain peers from all attestation subnets (most suitable for bootstrap nodes). -
Official docker images published at https://hub.docker.com/r/statusim/nimbus-eth2
-
Official Windows binaries created from a reproducible build recipe.
-
An option to enable the automatic updating of IP:Port in the ENR (off by default, specify
--enr-auto-update:true
to turn it on)
We've fixed:
-
A bug that had the potential to completely halt all syncing activity.
-
Inefficient processing of blocks with Eth1 deposits which occassionally led to increased latencies when delivering attestations.
-
Outdated records in our bootstrap nodes list.
-
An Eth1 syncing issue which manifested itself as a "Corrupted deposits history detected" error.
-
Non-standard encoding of certain data types such as signatures and bit sequences within the results of JSON-RPC requests.
We've deprecated:
-
make beacon_node
will no longer compile the beacon node. You'll need to runmake nimbus_beacon_node
from now on -
On monday we'll phase out the old
master
branch. If you're still building Nimbus frommaster
, please switch tostable
.
A release that brings reproducible precompiled binaries for ARM devices and significant performance improvements.
New features:
-
Reproducible build recipe for creating Nimbus Linux binaries intended for ARM devices.
-
Improved attestation subnet walking logic: this brings significant reductions in bandwidth usage and CPU load.
-
Better usage of the Sqlite3 checkpointing API (minor performance improvement).
-
Larger window for the candidate attestations included in blocks: this can lead to higher block rewards.
We've fixed:
- Incorrect
attnets
value announced in ENR records.
The 1.0.5 release was retracted because it included a potential optimization to the reproducible build scripts that turned out to create a buggy binary for AMD64 systems. Manually built binaries through the Makefiles were not affected. After fixing the problem, the release was re-published as 1.0.6 with the same release notes.
A release bringing further stability improvements and minor performance optimisations.
New features:
-
Nimbus can now be safely shut down with the SIGTERM signal on POSIX systems.
-
New discovery IP limits making theoretic eclipse attack much more costly.
-
A new
make benchmarks
target for obtaining a performance score for your system. -
Upgrade of the BLST library bringing minor performance improvement.
We've fixed:
-
Gossipsub resource leaks that may reduce the quality of the gossipsub mesh and reduce the attestation effectiveness of the client.
-
Incomplete validation of the forwarded attestations that may affect negatively the peer score of Nimbus.
-
An issue halting the activity of the Eth1 monitor.
-
The incorrect zero validator balance displayed while the node is syncing.
-
A regression preventing Nimbus to be used with custom testnet metadata files.
A release fixing issues that have contributed to Nimbus's lower peer scores on the network.
New features:
-
New metrics tracking the syncing progress of the Eth1 deposit contract monitor.
-
A new
web3 test
command for testing the compatibility of a web3 provider before using it.
We've fixed:
-
Incorrect timing when sending aggregated attestations.
-
Stale ENR records not taking into account the dynamic attestation subnet hopping.
-
An invalid error message produced by the
deposits exit
command (validator state unknown).
A release that fixes an issue regarding the occasional missed block proposal.
Proposing a block is arguably the most important duty you have as a validator. So it's important you update at your earliest convenience.
New features:
- 8 new JSON-RPC calls that bring us to feature parity with the official beacon node API.
We've fixed:
-
A deposit merkle proofs generation issue occasionally resulting in missed block proposals shortly after a new Eth1 head was selected.
-
Slow status bar updates in the absense of logging messages.
A release with a number of important fixes and optimisations.
Please update at your earliest convenience.
In order to minimise downtime, we recommend updating and rebuilding the beacon node before restarting.
New features:
- More conservative Eth1 syncing requests to reduce the likelihood of going over the maximum allowed burst rates under the Infura free plan (predominantly aimed at those running Nimbus on resource-restricted device like Raspberry Pi's) + more resiliency in case of errors.
We've fixed:
-
A "Only one concurrent read allowed" crash reported by multiple users.
-
An error in the default configuration preventing the node from discovering peers on mainnet unless the
--network=mainnet
flag was passed. Please note that this not affect users starting their node with the./run-mainnet-beacon-node
command. -
The fractional part of the ETH balance in the Nimbus status bar (the value displayed should now be correct).
-
An issue that occasionally caused the Eth1 syncing process to get stuck before reaching the head of the chain.
-
Unnecessary network traffic related to GossipSub
IHAVE
. -
Incorrect gossipsub pruning which occasionally resulted in messages getting lost.
-
An issue where an excessively long graffiti string could cause a crash on startup.
-
A Linux-only issue that resulted in the
deposits import
command ignoring its supplied arguments.
As promised, a slightly more polished release before Mainnet launch ✨
Please make sure you update to this release before Eth2 genesis this Tuesday (December 1 12:00:23 UTC), as it contains some important improvements.
New features:
-
Updated list of bootstrap nodes for Mainnet.
-
Prometheus metrics for validator balances. The beacon node will also display the total balance of all attached validators in the status footer by default.
-
deposits import
now automagically finds thevalidator_keys
directory produced by theeth2.0-deposit-cli
if it is located in the same working directory. -
A
deposits exit
command for submitting a voluntary validator exit. -
A
record
CLI command for inspecting and creating ENR records. -
An
--agent-string
option for specifying how Nimbus will present itself in LibP2P messages. The default value is nownimbus
. -
New RPC calls to track node and config status. Specifically, a JSON-RCP call for inspecting the active config preset (
get_v1_config_spec
).
We've fixed:
-
Inaccurate peer counts (an occasional mismatch between the number of syncing peers and GossipSub peers) -- the default peer limit has been increased to maintain a healthy gossip mesh.
-
High bandwidth usage of GossipSub (due to sub-optimal caching and lack of limits in the IWANT/IHAVE exchange messages) -- we're now using the latest spec GossipSub parameters.
-
High sync memory footprint -- we've reduced the number of sync workers from 20 to 10 (note, this should not affect sync speed).
We're happy to join the other client teams in announcing our v1.0.0
release
candidate with support for Mainnet ✨
You can use this release/binary to set up your rig and monitoring for Eth2 genesis next Tuesday (December 1 12:00:23 UTC).
N.B. There will be at least one more release, before December 1st. In particular, we are planning a more polished release for Sunday which will act as a drop-in replacement for this release candidate.
Don't worry if your peer count appears low at first -- It should increase as more validators connect to Mainnet.
Highlights include:
-
The addition of a deposit contract "state snapshot" to network metadata. This allows the client to skip syncing deposits made prior to the snapshot.
-
A much faster startup time. We've removed the deposits table from the database, which means the client no longer needs to process all deposits on start-up.
-
The Eth1 monitor no longer starts if the beacon node has zero validators attached to it.
-
The genesis detection code is now optional and disabled by default.
-
An RPC call to get Chronos futures at runtime.
-
Eth2 spec gossip parameters.
We've fixed:
-
A database corruption issue affecting Pyrmont nodes.
-
Unnecessary copy/memory alloc when loading DbSeq entries.
-
A block production issue affecting clients that hadn't finished downloading the latest deposits.
New features:
- New RPC APIs for inspecting the internal state of the Eth1 monitor.
We've fixed:
-
A fork-choice issue causing Nimbus to get stuck on a particular slot.
-
A logic error causing Nimbus to vote for an incorrect Eth1 block.
-
A crash during initialization when the web3 provider is refusing to serve data (e.g. due to exceeded request quota).
New features:
-
Support for the Pyrmont testnet.
-
The PCRE library is no longer necessary for building Nimbus.
-
Sensitive files such as keystores are now accessible only to the user of the beacon node on POSIX systems (the group rights have been dropped).
We've fixed:
-
An issue preventing blocks to be downloaded when the client goes out of sync.
-
Resource leaks that may lead to reduction of network activity due to a build-up of malfunctioning peer connections.
A bugfix release addressing issues discovered in the Toledo network.
New features:
-
GossipSub 1.1
-
The beacon node status bar (footer) now contains a time-left-until-synced estimate.
-
A JSON-RPC method
setLogLevel
for dynamically changing the log level of selected components at run-time. -
The ability to launch Nimbus with a partially-synced Geth node.
We've fixed:
-
A bug preventing the node from proposing blocks when connected to a web3 provider
-
An invalid "corrupted database" error message appearing on start-up
-
Incorrectly set message-ids in gossip message causing other clients to penalise and potentially disconnect our nodes from the network.
-
An issue occuring when Nimbus is paired with a Geth node that is not fully synced.
Nimbus eth2
0.6.0 was the first externally audited and stable release
of our beacon node software. When compared to the 0.5x series, it features
significant reductions in storage and memory requirements, a faster sync
speed, and a plethora of usability and security enhancements across the
board. Under normal network conditions, the delivery rate of attestations
and block proposals is expected to be above 99%. Going forward, our release
schedule will start to accelerate, with multiple new releases expected before
the Eth2 mainnet launch.
Changelog highlights include:
-
Full support for the 1.0 Eth2 phase0 spec and the monitoring of the mainnet validator deposit contract.
-
LibP2P and GossipSub fixes which drastically improve the delivery of attestations and blocks (nearly 100% expected rate of delivery).
-
Fixes for all major resource leaks: you no longer need to restart your node to improve its performance.
-
Efficient caching and storage mechanisms: ensures our memory consumption remains comparatively low both during smooth and turbulent network conditions.
-
Several storage and networking optimisations leading to an order of magnitude improvement in beacon chain sync speed.
-
Audits to our codebase by ConsenSys Diligence, NCC Group and Trail of Bits. More than 60 of the security findings have already been addressed. The remaining items will be resolved before mainnet launch.
-
Support for pairing with a locally running Geth instance to allow for decentralised monitoring of the validator deposit contract.
-
An extensive user guide for managing the beacon node.
-
Slashing protection mechanisms + database.
-
Support for storing the validator signing keys in a separate process, isolated from the network, with a minimal attack surface.