The historical view is not implemented for MemoryStore

[maschad]

As we integrated launchTestNode across the whole monorepo, instances of the the error

FuelError: got error during work with storage The historical view is not implemented for `MemoryStore`

appear more frequently in test runs, which affects the consistency of our tests. This is obviously due to the fact that we are using an in-memory database, but using a persistent key-value store comes with it's own challenges, and there aren't any test cases where we actually require historical data yet, so I wouldn't suggest that as a solution.

We should at the very least handle this error as it's not relevant to our test cases, so that we reduce flakiness in our CI checks.

Related to #2698

Related to #1755

[arboleya]

What is there to be fixed on the ts-sdk?

[maschad]

@arboleya the error is being thrown here on random tests, so I've tried a very rudimentary way to suppress it in the tests since it's not relevant to our test cases, but I haven't investigated the issue in depth to see the root cause.

[petertonysmith94]

Had a brief look and it appears to occur predominantly when we pass through a custom POA, however, this could be a red herring.

[arboleya]

Following @Torres-ssf investigations, I understood it's coming from fuel-core.

If that is true, it may have nothing to do with launchTestNode, which makes this issue a bit misleading.

[petertonysmith94]

Is this issue a P0?

[maschad]

@arboleya thanks for sharing that information, I'm still unsure how to reproduce the issue consistently. My primary rationale for raising this issue is that it's currently failing tests on #2811 and so I'll update the description according. At the very least we should handle the error / suppress it in test runs so it's not relevant to them.

@petertonysmith94 I've been experiencing recurring failing tests on #2811 and for that reason I've made it a p0

[maschad]

I've lowered this to a p2 after further investigation, by skipping the main test causing this issue, this error can be avoided relatively consistently, although it can still appear.

[Torres-ssf]

@maschad, I believe this error can be mitigated by using a longer --poa-interval-period when starting the node.

I am under the impression that this error tends to occur when the --poa-interval-period is set to a very short duration, such as a few milliseconds.

The reasoning is that a shorter --poa-interval-period causes nodes to be spawned more quickly, and since we are not using a persistent data solution, I suspect fuel-core eventually has trouble accessing data written in previous nodes, though I am not entirely sure about this.

[maschad]

@Torres-ssf I'm aware that the issue is related to --poa-interval-period as I had previously tried altering the period in previous commits but to no avail. Likewise I can't pinpoint the exact cause.

Unfortunately your recommendation is insufficient for this test use case (or any test similar to it) as it requires a transaction maturity > 1 and so we need a shorter interval period to increase block production time for such scenarios.

[maschad]

I'm unassigning myself from this for now as it would require substantial engineering effort and currently only one test would benefit from it. We can revisit this in the future if needs be.

[arboleya]

Should be automatically solved when fuel-core releases this:

• Don't use historical view during dry run fuel-core#2063

@maschad @Torres-ssf Please reopen if necessary.