## Issue Addressed Closes #2419 ## Proposed Changes Address a long-standing issue with the import of slashing protection data where the import would fail due to the data appearing slashable w.r.t the existing database. Importing is now idempotent, and will have no issues importing data that has been handed back and forth between different validator clients, or different implementations. The implementation works by updating the high and low watermarks if they need updating, and not attempting to check if the input is slashable w.r.t itself or the database. This is a strengthening of the minification that we started to do by default since #2380, and what Teku has been doing since the beginning. ## Additional Info The only feature we lose by doing this is the ability to do non-minified imports of clock drifted messages (cf. Prysm on Medalla). In theory, with the previous implementation we could import all the messages in case of clock drift and be aware of the "gap" between the real present time and the messages signed in the far future. _However_ for attestations this is close to useless, as the source epoch will advance as soon as justification occurs, which will require us to make slashable attestations with respect to our bogus attestation(s). E.g. if I sign an attestation 100=>200 when the current epoch is 101, then I won't be able to vote in any epochs prior to 101 becoming justified because 101=>102, 101=>103, etc are all surrounded by 100=>200. Seeing as signing attestations gets blocked almost immediately in this case regardless of our import behaviour, there's no point trying to handle it. For blocks the situation is more hopeful due to the lack of surrounds, but losing block proposals from validators who by definition can't attest doesn't seem like an issue (the other block proposers can pick up the slack).
8.7 KiB
Slashing Protection
The security of Ethereum 2.0's proof of stake protocol depends on penalties for misbehaviour, known
as slashings. Validators that sign conflicting messages (blocks or attestations), can be slashed
by other validators through the inclusion of a ProposerSlashing or AttesterSlashing on chain.
The Lighthouse validator client includes a mechanism to protect its validators against accidental slashing, known as the slashing protection database. This database records every block and attestation signed by validators, and the validator client uses this information to avoid signing any slashable messages.
Lighthouse's slashing protection database is an SQLite database located at
$datadir/validators/slashing_protection.sqlite which is locked exclusively when the validator
client is running. In normal operation, this database will be automatically created and utilized,
meaning that your validators are kept safe by default.
If you are seeing errors related to slashing protection, it's important that you act slowly and carefully to keep your validators safe. See the Troubleshooting section.
Initialization
The database will be automatically created, and your validators registered with it when:
- Importing keys from another source (e.g. Launchpad, Teku, Prysm,
ethdo). See the docs on importing keys. - Creating keys using Lighthouse itself (
lighthouse account validator create) - Creating keys via the validator client API.
Avoiding Slashing
The slashing protection database is designed to protect against many common causes of slashing, but is unable to prevent against some others.
Examples of circumstances where the slashing protection database is effective are:
- Accidentally running two validator clients on the same machine with the same datadir. The exclusive and transactional access to the database prevents the 2nd validator client from signing anything slashable (it won't even start).
- Deep re-orgs that cause the shuffling to change, prompting validators to re-attest in an epoch where they have already attested. The slashing protection checks all messages against the slashing conditions and will refuse to attest on the new chain until it is safe to do so (usually after one epoch).
- Importing keys and signing history from another client, where that history is complete. If you run another client and decide to switch to Lighthouse, you can export data from your client to be imported into Lighthouse's slashing protection database. See Import and Export.
- Misplacing
slashing_protection.sqliteduring a datadir change or migration between machines. By default Lighthouse will refuse to start if it finds validator keys that are not registered in the slashing protection database.
Examples where it is ineffective are:
- Running two validator client instances simultaneously. This could be two different clients (e.g. Lighthouse and Prysm) running on the same machine, two Lighthouse instances using different datadirs, or two clients on completely different machines (e.g. one on a cloud server and one running locally). You are responsible for ensuring that your validator keys are never running simultanously – the slashing protection DB cannot protect you in this case.
- Importing keys from another client without also importing voting history.
- If you use
--init-slashing-protectionto recreate a missing slashing protection database.
Import and Export
Lighthouse supports the slashing protection interchange format described in EIP-3076. An interchange file is a record of blocks and attestations signed by a set of validator keys – basically a portable slashing protection database!
With your validator client stopped, you can import a .json interchange file from another client
using this command:
lighthouse account validator slashing-protection import <my_interchange.json>
Instructions for exporting your existing client's database are out of scope for this document, please check the other client's documentation for instructions.
When importing an interchange file, you still need to import the validator keystores themselves separately, using the instructions for importing keystores into Lighthouse.
You can export Lighthouse's database for use with another client with this command:
lighthouse account validator slashing-protection export <lighthouse_interchange.json>
The validator client needs to be stopped in order to export, to guarantee that the data exported is up to date.
How Import Works
Since version 1.6.0 Lighthouse will ignore any slashable data in the import data and will safely update the low watermarks for blocks and attestations. It will store only the maximum-slot block for each validator, and the maximum source/target attestation. This is faster than importing all data while also being more resilient to repeated imports & stale data.
Minification
The exporter can be configured to minify (shrink) the data it exports by keeping only the
maximum-slot and maximum-epoch messages. Provide the --minify=true flag:
lighthouse account validator slashing-protection export --minify=true <lighthouse_interchange.json>
This may make the file faster to import into other clients, but is unnecessary for Lighthouse to Lighthouse transfers since v1.5.0.
Troubleshooting
Misplaced Slashing Database
If the slashing protection database cannot be found, it will manifest in an error like this:
Oct 12 14:41:26.415 CRIT Failed to start validator client reason: Failed to open slashing protection database: SQLError("Unable to open database: Error(Some(\"unable to open database file: /home/karlm/.lighthouse/mainnet/validators/slashing_protection.sqlite\"))").
Ensure that `slashing_protection.sqlite` is in "/home/karlm/.lighthouse/mainnet/validators" folder
Usually this indicates that during some manual intervention the slashing database has been misplaced. This error can also occur if you have upgraded from Lighthouse v0.2.x to v0.3.x without moving the slashing protection database. If you have imported your keys into a new node, you should never see this error (see Initialization).
The safest way to remedy this error is to find your old slashing protection database and move
it to the correct location. In our example that would be
~/.lighthouse/mainnet/validators/slashing_protection.sqlite. You can search for your old database
using a tool like find, fd, or your file manager's GUI. Ask on the Lighthouse Discord if you're
not sure.
If you are absolutely 100% sure that you need to recreate the missing database, you can start
the Lighthouse validator client with the --init-slashing-protection flag. This flag is incredibly
dangerous and should not be used lightly, and we strongly recommend you try finding
your old slashing protection database before using it. If you do decide to use it, you should
wait at least 1 epoch (~7 minutes) from when your validator client was last actively signing
messages. If you suspect your node experienced a clock drift issue you should wait
longer. Remember that the inactivity penalty for being offline for even a day or so
is approximately equal to the rewards earned in a day. You will get slashed if you use
--init-slashing-protection incorrectly.
Slashable Attestations and Re-orgs
Sometimes a re-org can cause the validator client to attempt to sign something slashable, in which case it will be blocked by slashing protection, resulting in a log like this:
Sep 29 15:15:05.303 CRIT Not signing slashable attestation error: InvalidAttestation(DoubleVote(SignedAttestation { source_epoch: Epoch(0), target_epoch: Epoch(30), signing_root: 0x0c17be1f233b20341837ff183d21908cce73f22f86d5298c09401c6f37225f8a })), attestation: AttestationData { slot: Slot(974), index: 0, beacon_block_root: 0xa86a93ed808f96eb81a0cd7f46e3b3612cafe4bd0367aaf74e0563d82729e2dc, source: Checkpoint { epoch: Epoch(0), root: 0x0000000000000000000000000000000000000000000000000000000000000000 }, target: Checkpoint { epoch: Epoch(30), root: 0xcbe6901c0701a89e4cf508cfe1da2bb02805acfdfe4c39047a66052e2f1bb614 } }
This log is still marked as CRIT because in general it should occur only very rarely,
and could indicate a serious error or misconfiguration (see Avoiding Slashing).
Limitation of Liability
The Lighthouse developers do not guarantee the perfect functioning of this software, or accept liability for any losses suffered. For more information see the Lighthouse license.