Attestant's Vouch & Dirk in Docker Compose, to be used with Eth Docker
This repo assumes a 2 vouch and 5 dirk setup with 3/5 threshold signing, and 3 Ethereum CL:EL full nodes.
cp default.env .env, edit it to choose whether to run vouch and dirk or just dirk or just vouch, set the instance ID
for dirk/vouch, and adjust host and domain names to fit your environment.
If you are going to use Vouch traces, set TRACE_URL and VOUCH_SAN in .env. Either copy tempo_authority.crt into
config/certs, or if you don't have a CA and server key for Tempo yet, let the next step create them.
./create-dirk-config.sh is meant to be run just once. It has no sanity checks and creates the CA, certs, and
config yml files. The config directory would then be copied to each server where a vouch or dirk instance runs. The
script gets names from .env, which must exist.
./ethd up to start Vouch/Dirk services.
On each of the Dirk instances, run docker compose run --rm create-wallet once.
Run ./ethd restart dirk to ensure Dirk loads this new wallet correctly.
Note key generation takes ~ 3.5 GiB of RAM
To create keys, adjust start and stop index in .env and then run docker compose run --rm create-accounts.
To verify the first of these keys for correctness, run docker compose run --rm verify-account
To create deposit data, using the same start and stop index in .env, run docker compose run --rm create-depositdata.
Adjust FORK_VERSION in .env if you are going to generate for a testnet.
You can then create a single deposit.json with:
jq -n '[inputs|add]' config/depositdata/deposit-val-{1..10}.json > ~/deposits.json
adjusting the {1..10} for the range you want to have in the file.
2 Vouch (one cold standby) and a 3/5 Dirk (3 threshold, 5 total) were chosen carefully. With 2 Vouch and 2/4 Dirk there would be a risk of slashing; 3 Vouch and 3/5 Dirk, Vouch might not get to threshold and never be able to sign duties.
1 Vouch and 2/3 Dirk would also work just as well.
This repository was created for a cross-region setup, with five hosts for Vouch/Dirk and 3 separate hosts for the CL:EL Ethereum nodes. It'd need some adjustment to be more flexible in these numbers.
An alternate setup could run 1 Vouch and a 2/3 Dirk threshold setup inside a single k8s cluster. This repo does not aim to support that use case.
The reason a cross-region setup was chosen is that while region outages are rare, they do occur. It is desirable for a staking node operator to be able to regain liveness even when an entire region fails.
With multiple Vouch instances, a degradation in the number of Dirk instances can result in the inability to sign. At its simplest, if there are 2 Vouch and 4 Dirk (out of originally 5) with a threshold of 3, then it is possible for one Vouch instance to obtain 2 signatures and the second Vouch instance to obtain 2 signatures, with neither reaching the threshold and no signature generated. If MEV is in use, the CL needs to "point back to" Vouch, and this as well requires the use of a single Vouch instance.
For this reason, it is recommended to run Vouch in container orchestration with cross-AZ failover, and have the second Vouch instance ready in case there is an outage for an entire region. Vouch is stateless and will start in seconds.
Slashing protection works like this:
- Vouch will ask Dirk for a threshold signature
- A Dirk that has already participated in one will refuse to do so again because of the slashing protection DB. That means if both Vouch are running simultaneously, one will get to at least 3/5, the other to at most 2/5 and won't get a full signature
- The slashing protection DB is kept locally by each Dirk
The recommendation by attestant.io is to restart Dirk instances after adding or removing keys, because of the way caching works.
To back up the wallet on each local Dirk instance, run docker compose run --rm export-keys and then save the resulting
file and the passphrase you used.
You will need to run this on all five (5) Dirk instances individually.
Each Dirk instance is its own entity. If a Dirk instance fails, the backup of that instance can be restored, and once it is rebuilt it will continue in the cluster.
By adding :prometheus.yml to COMPOSE_FILE in .env you can run a Prometheus that can remote-write to your Mimir or
Thanos, with the remote-write section in prometheus/custom-prom.yml.
In case you need prometheus to also scrape extra custom endpoints, add a yaml file in prometheus/file_sd folder. Example below
- labels:
job: JOB_LABEL
targets:
- JOB_ENDPOINT:JOB_PORTHuge THANK YOU to Jeff Schroeder at Jump Crypto for generously sharing his knowledge of this setup, and to Jim McDonald at attestant.io for creating these tools in the first place, and always having patience and time for explanations.
This is Vouchdirk Docker v1.2.1