[Merged by Bors] - Hotfix for API queries for activations

[lrettig]

Modify MeshService.LayersQuery to return ONLY the single, canonical block for each queried layer. Fixes performance issues for now by not querying any activations, proposals, etc.

Modify MeshService.AccountMeshDataQuery not to query activations to prevent memory/CPU bomb

Motivation

Closes #5315
Closes #5006

Changes

• Add method to mesh to return single, valid, applied block for a given layer
• Modify MeshService.LayersQuery to use this new method
• Additionally, remove query for activations (leave this empty for now)
• Modify MeshService.AccountMeshDataQuery not to query activations

Test Plan

Tests updated

TODO

• Explain motivation or link existing issue(s)
• Test changes and document test plan
• Update documentation as needed
• Update changelog as needed

[codecov]

Codecov Report

Attention: 6 lines in your changes are missing coverage. Please review.

Comparison is base (2e7d09e) 77.2% compared to head (6fe005f) 77.7%.
Report is 17 commits behind head on develop.

Files	Patch %	Lines
mesh/mesh.go	63.6%	4 Missing ⚠️
api/grpcserver/mesh_service.go	93.7%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff            @@
##           develop   #5334     +/-   ##
=========================================
+ Coverage     77.2%   77.7%   +0.4%     
=========================================
  Files          262     265      +3     
  Lines        30681   30862    +181     
=========================================
+ Hits         23716   23987    +271     
+ Misses        5462    5362    -100     
- Partials      1503    1513     +10     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[fasmat]

NIT: The named returns are not actually used (every return statement explicitly states what values are returned). If the names here are removed you can use err instead of err2 everywhere in this function 🙂

[lrettig]

yea, oof, i like named returns so I added this, but our strict linting rules no longer allow this =\

[fasmat]

NIT: I know this extends MeshAPIMock but would it make sense to instead use a gomock Mock here instead? These handwritten mocks are really hard to maintain, because changing anything usually breaks several tests at once because they are shared between them.

[lrettig]

Thanks for pointing this out. If we were doing this from scratch I'd agree, but the methods are all already mocked here:

go-spacemesh/api/grpcserver/grpcserver_test.go

Lines 247 to 321 in a05776e

	type MeshAPIMock struct{}
	// latest layer received.
	func (m *MeshAPIMock) LatestLayer() types.LayerID {
	return layerLatest
	}
	// latest layer approved/confirmed/applied to state
	// The real logic here is a bit more complicated, as it depends whether the node
	// is syncing or not. If it's not syncing, layers are applied to state as they're
	// verified by Hare. If it's syncing, Hare is not run, and they are applied to
	// state as they're confirmed by Tortoise and it advances pbase. This is all in
	// flux right now so keep this simple for the purposes of testing.
	func (m *MeshAPIMock) LatestLayerInState() types.LayerID {
	return layerVerified
	}
	func (m *MeshAPIMock) ProcessedLayer() types.LayerID {
	return layerVerified
	}
	func (m *MeshAPIMock) GetRewardsByCoinbase(types.Address) (rewards []*types.Reward, err error) {
	return []*types.Reward{
	{
	Layer: layerFirst,
	TotalReward: rewardAmount,
	LayerReward: rewardAmount,
	Coinbase: addr1,
	SmesherID: rewardSmesherID,
	},
	}, nil
	}
	func (m *MeshAPIMock) GetRewardsBySmesherId(types.NodeID) (rewards []*types.Reward, err error) {
	return []*types.Reward{
	{
	Layer: layerFirst,
	TotalReward: rewardAmount,
	LayerReward: rewardAmount,
	Coinbase: addr1,
	SmesherID: rewardSmesherID,
	},
	}, nil
	}
	func (m *MeshAPIMock) GetLayer(tid types.LayerID) (*types.Layer, error) {
	if tid.After(layerCurrent) {
	return nil, errors.New("requested layer later than current layer")
	} else if tid.After(m.LatestLayer()) {
	return nil, errors.New("haven't received that layer yet")
	}
	ballots := []*types.Ballot{ballot1}
	blocks := []*types.Block{block1, block2, block3}
	return types.NewExistingLayer(tid, ballots, blocks), nil
	}
	func (m *MeshAPIMock) GetLayerVerified(tid types.LayerID) (*types.Block, error) {
	return block1, nil
	}
	func (m *MeshAPIMock) GetATXs(
	context.Context,
	[]types.ATXID,
	) (map[types.ATXID]*types.VerifiedActivationTx, []types.ATXID) {
	atxs := map[types.ATXID]*types.VerifiedActivationTx{
	globalAtx.ID(): globalAtx,
	globalAtx2.ID(): globalAtx2,
	}
	return atxs, nil
	}
	func (m *MeshAPIMock) MeshHash(types.LayerID) (types.Hash32, error) {
	return types.RandomHash(), nil
	}

. I don't think it makes sense to fully rewrite this at this stage, and I don't see much benefit from basing the existing MeshAPIMock type on the auto-generated MockmeshAPI type. let me know if you disagree.

[lrettig]

bors merge

[spacemesh-bors]

Build failed:

• systest-status

[kacpersaw]

bors merge

[spacemesh-bors]

Build failed:

• systest-status

[bobbychen]

Why this hotfix haven't merge into this release~~~,I think it is a problem, hang on long time when query a layer detail.

[kacpersaw]

bors merge

[spacemesh-bors]

Build failed:

• systest-status

[kacpersaw]

bors merge

[spacemesh-bors]

Pull request successfully merged into develop.

Build succeeded:

• ci-status
• systest-status