Oracle daemon for Lido decentralized staking service: Monitoring the state of the protocol across both layers and submitting regular update reports to the Lido smart contracts.
There are 3 modules in the oracle:
- Accounting
- Ejector
- CSM
Accounting module updates the protocol TVL, distributes node-operator rewards, updates information about the number of exited and stuck validators and processes user withdrawal requests. Also Accounting module makes decision to turn on/off the bunker.
Flow
The oracle work is delineated by time periods called frames. Oracles finalize a report in each frame. The default Accounting Oracle frame length on mainnet is 225 epochs, which is 24 hours (it could be changed by DAO). The frame includes these stages:
- Waiting - oracle starts as daemon and wakes up every 12 seconds (by default) in order to find the last finalized slot (ref slot). If ref slot missed, Oracle tries to find previous non-missed slot.
- Data collection: oracles monitor the state of both the execution and consensus layers and collect the data;
- Hash consensus: oracles analyze the data, compile the report and submit its hash to the HashConsensus smart contract;
- Core update report: once the quorum of hashes is reached, meaning required number of Oracles submitted the same hash, one of the oracles chosen in turn submits the actual report to the AccountingOracle contract, which triggers the core protocol state update, including the token rebase, finalization of withdrawal requests, and deciding whether to go in the bunker mode.
- Extra data report: an third phase report carrying additional information. It can be delivered multi-transactionally. Delivers stuck and exited validators count by node operators. For some modules unlocks rewards distribution.
Ejector module requests Lido validators to eject via events in Execution Layer when the protocol requires additional funds to process user withdrawals.
Flow
- Finds out how much ETH is needed to cover withdrawals.
- Predicts mean Lido income into Withdrawal and Execution Rewards Vaults.
- Figures out when the next validator will be withdrawn.
- Form a validator's queue with enough validators to fill all withdrawals requests.
- Force eject validators from Node Operator with boosted exits flag even no withdrawal requests.
- Encode validators data and send transaction
Only Oracle:
- vCPUs - 2
- Memory - 8 GB
Oracle + KAPI:
- vCPU - 4
- Memory - 16 GB
To prepare the report, Oracle fetches up to 10 days old events, makes historical requests for balance data and makes simulated reports on historical blocks. This requires an archive execution node. Oracle needs two weeks of archived data.
Client | Tested | Notes |
---|---|---|
Geth | π’ | --gcmode=archive --syncmode=snap OR --gcmode=archive --syncmode=full |
Nethermind | π΄ | Not tested yet |
Besu | π’ | Use --rpc-max-logs-range=100000 --sync-mode=FULL --data-storage-format="FOREST" --pruning-enabled --pruning-blocks-retained=100000 params |
Erigon | π’ | Use --prune=htc --prune.h.before=100000 --prune.t.before=100000 --prune.c.before=100000 params |
Also, to calculate some metrics for bunker mode Oracle needs archive consensus node.
Client | Tested | Notes |
---|---|---|
Lighthouse | π’ | Use --reconstruct-historic-states param |
Lodestar | π΄ | Not tested yet |
Nimbus | π΄ | Not tested yet |
Prysm | π’ | Use --grpc-max-msg-size=104857600 --enable-historical-state-representation=true --slots-per-archive-point=1024 params |
Teku | π’ | Use --data-storage-mode=archive --data-storage-archive-frequency=1024 --reconstruct-historic-states=true params |
This is a separate service that uses Consensus and Execution Clients to fetch all lido keys. It stores the latest state of lido keys in database.
Oracle daemon could be run using a docker container. Images is available on Docker Hub. Pull the image using the following command:
docker pull lidofinance/oracle:{tag}
Where {tag}
is a version of the image. You can find the latest version in the releases
OR
You can build it locally using the following command (make sure build it from latest release):
docker build -t lidofinance/oracle .
Full variables list could be found here.
- Use .env.example file content to create your own
.env
file. Set required values. It will be enough to run the oracle in check mode. - Check that your environment is ready to run the oracle using the following command:
If everything is ok, you will see that all required checks are passed and your environment is ready to run the oracle.
docker run -ti --env-file .env --rm lidofinance/oracle:{tag} check
-
By default, the oracle runs in dry mode. It means that it will not send any transactions to the Ethereum network. To run Oracle in production mode, set
MEMBER_PRIV_KEY
orMEMBER_PRIV_KEY_FILE
environment variable:MEMBER_PRIV_KEY={value}
Where
{value}
is a private key of the Oracle member account or:MEMBER_PRIV_KEY_FILE={path}
Where
{path}
is a path to the private key of the Oracle member account. -
Run the container using the following command:
docker run --env-file .env lidofinance/oracle:{tag} {type}
Replace
{tag}
with the image version and{type}
with one of the two types of oracles: accounting or ejector.
Note: of course, you can pass env variables without using
.env
file. For example, you can run the container using the following command:docker run --env EXECUTION_CLIENT_URI={value} --env CONSENSUS_CLIENT_URI={value} --env KEYS_API_URI={value} --env LIDO_LOCATOR_ADDRESS={value} lidofinance/oracle:{tag} {type}
Oracle could be executed once in "manual" mode. To do this setup DAEMON
variable to 'False'.
Note: Use -it
option to run manual mode in Docker container in interactive mode.
Example docker run -ti --env-file .env --rm lidofinance/oracle:{tag} {type}
In this mode Oracle will build report as usual (if contracts are reportable) and before submitting transactions Oracle will ask for manual input to send transaction.
In manual mode all sleeps are disabled and ALLOW_REPORTING_IN_BUNKER_MODE
is True.
Name | Description | Required | Example value |
---|---|---|---|
EXECUTION_CLIENT_URI |
URI of the Execution Layer client | True | http://localhost:8545 |
CONSENSUS_CLIENT_URI |
URI of the Consensus Layer client | True | http://localhost:5052 |
KEYS_API_URI |
URI of the Keys API | True | http://localhost:8080 |
LIDO_LOCATOR_ADDRESS |
Address of the Lido contract | True | 0x1... |
CSM_MODULE_ADDRESS |
Address of the CSModule contract | CSM only | 0x1... |
MEMBER_PRIV_KEY |
Private key of the Oracle member account | False | 0x1... |
MEMBER_PRIV_KEY_FILE |
A path to the file contained the private key of the Oracle member account. It takes precedence over MEMBER_PRIV_KEY |
False | /app/private_key |
GW3_ACCESS_KEY |
An access key to gw3.io IPFS provider | CSM only | 123456789-1234-5678-9012-123456789012 |
GW3_ACCESS_KEY_FILE |
A path to a file with an access key to gw3.io IPFS provider | CSM only | /app/gwt_access |
GW3_SECRET_KEY |
A secret key to gw3.io IPFS provider | CSM only | aBcD1234... |
GW3_SECRET_KEY_FILE |
A path to a file with a secret key to gw3.io IPFS provider | CSM only | /app/gwt_secret |
PINATA_JWT |
JWT token to access pinata.cloud IPFS provider | CSM only | aBcD1234... |
PINATA_JWT_FILE |
A path to a file with a JWT token to access pinata.cloud IPFS provider | CSM only | /app/pintata_secret |
FINALIZATION_BATCH_MAX_REQUEST_COUNT |
The size of the batch to be finalized per request (The larger the batch size, the more memory of the contract is used but the fewer requests are needed) | False | 1000 |
EL_REQUESTS_BATCH_SIZE |
The amount of entities that would be fetched in one request to EL | False | 1000 |
ALLOW_REPORTING_IN_BUNKER_MODE |
Allow the Oracle to do report if bunker mode is active | False | True |
DAEMON |
If False Oracle runs one cycle and ask for manual input to send report. | False | True |
TX_GAS_ADDITION |
Used to modify gas parameter that used in transaction. (gas = estimated_gas + TX_GAS_ADDITION) | False | 100000 |
CYCLE_SLEEP_IN_SECONDS |
The time between cycles of the oracle's activity | False | 12 |
MAX_CYCLE_LIFETIME_IN_SECONDS |
The maximum time for a cycle to continue | False | 3000 |
SUBMIT_DATA_DELAY_IN_SLOTS |
The difference in slots between submit data transactions from Oracles. It is used to prevent simultaneous sending of transactions and, as a result, transactions revert. | False | 6 |
HTTP_REQUEST_TIMEOUT_EXECUTION |
Timeout for HTTP execution layer requests | False | 120 |
HTTP_REQUEST_TIMEOUT_CONSENSUS |
Timeout for HTTP consensus layer requests | False | 300 |
HTTP_REQUEST_RETRY_COUNT_CONSENSUS |
Total number of retries to fetch data from endpoint for consensus layer requests | False | 5 |
HTTP_REQUEST_SLEEP_BEFORE_RETRY_IN_SECONDS_CONSENSUS |
The delay http provider sleeps if API is stuck for consensus layer | False | 12 |
HTTP_REQUEST_TIMEOUT_KEYS_API |
Timeout for HTTP keys api requests | False | 120 |
HTTP_REQUEST_RETRY_COUNT_KEYS_API |
Total number of retries to fetch data from endpoint for keys api requests | False | 300 |
HTTP_REQUEST_SLEEP_BEFORE_RETRY_IN_SECONDS_KEYS_API |
The delay http provider sleeps if API is stuck for keys api | False | 300 |
HTTP_REQUEST_TIMEOUT_IPFS |
Timeout for HTTP requests to an IPFS provider | False | 30 |
HTTP_REQUEST_RETRY_COUNT_IPFS |
Total number of retries to fetch data from an IPFS provider | False | 3 |
EVENTS_SEARCH_STEP |
Maximum length of a range for eth_getLogs method calls | False | 10000 |
PRIORITY_FEE_PERCENTILE |
Priority fee percentile from prev block that would be used to send tx | False | 3 |
MIN_PRIORITY_FEE |
Min priority fee that would be used to send tx | False | 50000000 |
MAX_PRIORITY_FEE |
Max priority fee that would be used to send tx | False | 100000000000 |
CSM_ORACLE_MAX_CONCURRENCY |
Max count of dedicated workers for CSM module | False | 2 |
CACHE_PATH |
Directory to store cache for CSM module | False | . |
LIDO_LOCATOR_ADDRESS=0xC1d0b3DE6792Bf6b4b37EccdcC24e45978Cfd2Eb ALLOW_REPORTING_IN_BUNKER_MODE=False
A few basic alerts, which can be configured in the Prometheus Alertmanager.
groups:
- name: oracle-alerts
rules:
- alert: AccountBalance
expr: lido_oracle_account_balance / 10^18 < 1
for: 10m
labels:
severity: critical
annotations:
summary: "Dangerously low account balance"
description: "Account balance is less than 1 ETH. Address: {.labels.address}: {.value} ETH"
- alert: OutdatedData
expr: (lido_oracle_genesis_time + ignoring (state) lido_oracle_slot_number{state="head"} * 12) < time() - 300
for: 1h
labels:
severity: critical
annotations:
summary: "Outdated Consensus Layer HEAD slot"
description: "Processed by Oracle HEAD slot {.value} too old"
Note: all metrics are prefixed with
lido_oracle_
by default.
The oracle exposes the following basic metrics:
Metric name | Description | Labels |
---|---|---|
build_info | Build info | version, branch, commit |
env_variables_info | Env variables for the app | ACCOUNT, LIDO_LOCATOR_ADDRESS, CSM_MODULE_ADDRESS, FINALIZATION_BATCH_MAX_REQUEST_COUNT, EL_REQUESTS_BATCH_SIZE, MAX_CYCLE_LIFETIME_IN_SECONDS |
genesis_time | Fetched genesis time from node | |
account_balance | Fetched account balance from EL | address |
slot_number | Last fetched slot number from CL | state (head or finalized ) |
block_number | Last fetched block number from CL | state (head or finalized ) |
functions_duration | Histogram metric with duration of each main function in the app | name, status |
el_requests_duration | Histogram metric with duration of each EL request | endpoint, call_method, call_to, code, domain |
cl_requests_duration | Histogram metric with duration of each CL request | endpoint, code, domain |
keys_api_requests_duration | Histogram metric with duration of each KeysAPI request | endpoint, code, domain |
keys_api_latest_blocknumber | Latest block number from KeysAPI metadata | |
transactions_count | Total count of transactions. Success or failure | status |
member_info | Oracle member info | is_report_member, is_submit_member, is_fast_lane |
member_last_report_ref_slot | Member last report ref slot | |
frame_current_ref_slot | Current frame ref slot | |
frame_deadline_slot | Current frame deadline slot | |
frame_prev_report_ref_slot | Previous report ref slot | |
contract_on_pause | Contract on pause |
Special metrics for accounting oracle:
Metric name | Description | Labels |
---|---|---|
accounting_is_bunker | Is bunker mode enabled | |
accounting_cl_balance_gwei | Reported CL balance in gwei | |
accounting_el_rewards_vault_wei | Reported EL rewards in wei | |
accounting_withdrawal_vault_balance_wei | Reported withdrawal vault balance in wei | |
accounting_exited_validators | Reported exited validators count for each operator | module_id, no_id |
accounting_stuck_validators | Reported stuck validators count for each operator | module_id, no_id |
accounting_delayed_validators | Reported delayed validators count for each operator | module_id, no_id |
Special metrics for ejector oracle:
Metric name | Description | Labels |
---|---|---|
ejector_withdrawal_wei_amount | To withdraw amount | |
ejector_max_withdrawal_epoch | Max withdrawal epoch among all Lido validators on CL | |
ejector_validators_count_to_eject | Validators count to eject |
Special metrics for CSM oracle:
Metric name | Description | Labels |
---|---|---|
TBD | TBD |
Python version: 3.12
- Setup poetry
- Install dependencies
poetry install
Required variables for accounting and ejector modules
export EXECUTION_CLIENT_URI=...
export CONSENSUS_CLIENT_URI=...
export KEYS_API_URI=...
export LIDO_LOCATOR_ADDRESS=...
Required variables for CSM module
export EXECUTION_CLIENT_URI=...
export CONSENSUS_CLIENT_URI=...
export KEYS_API_URI=...
export LIDO_LOCATOR_ADDRESS=...
export CSM_MODULE_ADDRESS=...
export MAX_CYCLE_LIFETIME_IN_SECONDS=60000 # Reasonable high value to make sure the oracle has enough time to process the whole frame.
Run oracle module
poetry run python -m src.main {module}
Where {module}
is one of:
accounting
ejector
csm
check
poetry run pytest .
Used the following tools:
- black
- pylint
- mypy See the configuration for details for each linter.
Make sure that your code is formatted correctly and passes all checks:
black tests
pylint src tests
mypy src
2023 Lido [email protected]
This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, version 3 of the License, or any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program. If not, see https://www.gnu.org/licenses/.
To create new release:
- Merge all changes to the
master
branch - Navigate to Repo => Actions
- Run action "Prepare release" action against
master
branch - When action execution is finished, navigate to Repo => Pull requests
- Find pull request named "chore(release): X.X.X" review and merge it with "Rebase and merge" (or "Squash and merge")
- After merge release action will be triggered automatically
- Navigate to Repo => Actions and see last actions logs for further details