db-sync fails to restart with `Error: Shelley.validateGenesisDistribution`

[mkoura]

OS
Fedora 34

Versions
12.0.0-pre2 and current master: 15a434be9df084d04b8d2bf18df833470befecc8
PostgreSQL version: 11.11

Build/Install Method
The method you use to build or install cardano-db-sync: nix

Additional context
The cardano node is on latest master, Alonzo era, Protocol Version 6.

Problem Report
Db-sync is running without any issues until restarted. After restart it fails to start with
stdout:

[db-sync-node:Info:6] [2021-10-08 14:20:14.56 UTC] Schema migration files validated
[db-sync-node:Info:6] [2021-10-08 14:20:14.56 UTC] Running database migrations
[db-sync-node:Info:6] [2021-10-08 14:20:14.81 UTC] Using byron genesis file from: "./state-cluster0/byron/genesis.json"
[db-sync-node:Info:6] [2021-10-08 14:20:14.81 UTC] Using shelley genesis file from: "./state-cluster0/shelley/genesis.json"
[db-sync-node:Info:6] [2021-10-08 14:20:14.81 UTC] Using alonzo genesis file from: "./state-cluster0/shelley/genesis.alonzo.json"
[db-sync-node:Info:6] [2021-10-08 14:20:14.82 UTC] NetworkMagic: 42
[db-sync-node:Info:6] [2021-10-08 14:20:14.82 UTC] Initial genesis distribution present and correct
[db-sync-node:Info:6] [2021-10-08 14:20:14.82 UTC] Total genesis supply of Ada: 45000000000.000000
[db-sync-node:Info:6] [2021-10-08 14:20:14.82 UTC] Validating Genesis distribution

stderr:

Error: Shelley.validateGenesisDistribution: Expected initial block to have 1 but got 2

Might be related to #766 however my local network has

"TestShelleyHardForkAtEpoch": 1

(not "0").

It is a regression, with the same node and db-sync configuration it is possible to restart db-sync 11.0.4 with no issue.

[erikd]

Where is the config you are using?

[mkoura]

node config
dbsync config

[kderme]

@mkoura Could we also have the genesis files?

[mkoura]

@kderme
Shelley
Alonzo
Byron

The shelley-related genesis files above are generated from this genesis.spec.json using cardano-cli genesis create ... and the byron genesis is generated using cardano-cli byron genesis genesis ....

[kderme]

I think that the initialFunds and staking are only used for testnets that fork directly to Shelley, that is "TestShelleyHardForkAtEpoch": 0. So this must be an issue with the generator.

Note that previous version 11.0.4 would ignore these fields.

[kderme]

I found this comment in the consensus code:

-- NOTE: the initial staking and funds in the 'ShelleyGenesis' are ignored,
-- /unless/ configured to skip the Byron era and hard fork to Shelley or a later
-- era from the start using @TriggerHardForkAtEpoch 0@ for testing purposes.

So it seems the node ignores these fields. I would suggest that db-sync does not support these kind of configuration and it requires "TestShelleyHardForkAtEpoch": 0 when there are initial staking or funds.

[erikd]

We should detect this invalid configuration (TestShelleyHardForkAtEpoch /= 0 but with a Shelley staking and funds defined) and panicAbort on it, with a good error message.

[erikd]

Fixed in #896 .