Apparent performance could be wrong for brief times (because of multiple sources of truth)

[Anviking]

Context

The Pool.Metrics chain-follower ensures a source of stake distributions and pool productions that are consistent with each other (always on the same fork).

The performance calculation is:
Apparent Performance = (n / N) * (S / s)
where

• n is the number of blocks produced by the pool in the last epoch
• N is the number of slots in the epoch
• s is the stake owned by the pool in the last epoch
• S is the total stake delegated to pools in the last epoch

In the implementation however,
https://github.com/input-output-hk/cardano-wallet/blob/master/lib/core/src/Cardano/Pool/Metrics.hs#L239

the tip used to calculate N for the current epoch is retrieved separately from the node.

Because it is a different source of data from the DB, it is not guaranteed to be on the same chain as the DB.

Information	-
Version
Platform
Installation

Steps to Reproduce

Not tested:

Consider the scenario:

1. Slots per epoch = 20, poolA has 100% stake
2. stake-pool-worker syncs to slot 2.3
3. node rolls forward to 2.19
4. user calls listStakePools without the worker yet noticing the roll-foward

Expected behavior

Performance is calculated from consistent sources of data.

Likely we'd expect:
For epoch 2:
s / S = 1
N = 3
n = 3

AP = 100%

Actual behavior

Hypothetical, not tested:

For epoch 2:
s / S = 1
N = 19
n = 3

AP = 16%

(As the chain follower catches up, the inconsistency would disappear)

Resolution

---

We could calculate the N using the tip from the Pool DB instead of the tip from the node's REST api. ==> #1168

QA

• Added some unit test which checks that the tip used for the performance calculation (which happen to be the reported when the pools aren't in sync) is the right one, and not the network's one: https://github.com/input-output-hk/cardano-wallet/blob/master/lib/core/test/unit/Cardano/Pool/MetricsSpec.hs#L153-L155