Stateless observation hydra node #1196

v0d1ch · 2023-12-05T09:25:32Z

TODO:

Make E2E tests green with multiple heads on a network
fromPlutusCurrencySymbol partial
FIXME in broken Greetings
FIXME on comparing observations in HandlersSpec.hs
Test observe functions with multiple Head UTxO
Actually get rid of Hydra.Chain.Direct.State module from production code (use it only to generate transactions and rename it)

CHANGELOG updated or not needed
Documentation updated or not needed
Haddocks updated or not needed
No new TODOs introduced or explained herafter

github-actions · 2023-12-05T10:34:15Z

Transactions Costs

Sizes and execution budgets for Hydra protocol transactions. Note that unlisted parameters are currently using arbitrary values and results are not fully deterministic and comparable to previous runs.

Metadata
Generated at	2023-12-08 21:01:07.763378916 UTC
Max. memory units	14000000
Max. CPU units	10000000000
Max. tx size (kB)	16384

Script summary

Name	Hash	Size (Bytes)
νInitial	985245919fcc6c0c5cd116023cd2c947c43e80dcbb5075fe12433fbb	4072
νCommit	7cb20fa71eb4c563ca283566ebe0aa65859d96c3f8cba35c52c181fd	2043
νHead	7a36661f5c15e9f1783aeaab890812c59b7286cbbc6de762d3110772	8816
μHead	8b111ac12274e46314769295a1c5dcab1d260096fc469fd698065463*	3851

The minting policy hash is only usable for comparison. As the script is parameterized, the actual script is unique per Head.

Cost of Init Transaction

Parties	Tx size	% max Mem	% max CPU	Min fee ₳
1	4379	10.44	4.03	0.46
2	4576	12.55	4.83	0.49
3	4777	14.89	5.72	0.52
5	5180	19.22	7.35	0.59
10	6184	30.24	11.53	0.75
41	12422	99.15	37.65	1.77

Cost of Commit Transaction

This is using ada-only outputs for better comparability.

UTxO	Tx size	% max Mem	% max CPU	Min fee ₳
1	534	11.37	4.44	0.30
2	721	15.04	6.07	0.35
3	911	18.85	7.75	0.40
5	1282	26.90	11.27	0.51
10	2214	49.55	20.97	0.80
19	3904	99.43	41.75	1.43

Cost of CollectCom Transaction

Parties	UTxO (bytes)	Tx size	% max Mem	% max CPU	Min fee ₳
1	57	484	21.12	8.27	0.40
2	114	590	32.31	12.77	0.53
3	169	700	46.93	18.64	0.69
4	227	810	62.94	25.15	0.88
5	281	920	74.51	30.12	1.01
6	337	1031	94.84	38.41	1.24

Cost of Close Transaction

Parties	Tx size	% max Mem	% max CPU	Min fee ₳
1	541	16.68	7.67	0.36
2	765	18.45	9.34	0.40
3	907	19.93	10.76	0.43
5	1270	23.61	14.06	0.50
10	2141	32.24	22.02	0.66
50	8676	96.29	82.31	1.90

Cost of Contest Transaction

Parties	Tx size	% max Mem	% max CPU	Min fee ₳
1	576	20.22	8.96	0.40
2	738	21.93	10.47	0.44
3	924	24.28	12.40	0.48
5	1301	28.20	15.80	0.55
10	2182	37.83	24.05	0.73
43	7992	98.89	77.51	1.86

Cost of Abort Transaction

Some variation because of random mixture of still initial and already committed outputs.

Parties	Tx size	% max Mem	% max CPU	Min fee ₳
1	4309	18.98	8.16	0.55
2	4471	32.06	13.96	0.71
3	4559	42.29	18.31	0.82
4	4842	66.64	29.32	1.11
5	4917	80.70	35.40	1.27

Cost of FanOut Transaction

Involves spending head output and burning head tokens. Uses ada-only UTxO for better comparability.

Parties	UTxO	UTxO (bytes)	Tx size	% max Mem	% max CPU	Min fee ₳
5	0	0	4205	8.26	3.46	0.43
5	1	57	4240	9.28	4.13	0.44
5	5	285	4375	14.40	7.27	0.51
5	10	570	4546	20.34	10.99	0.59
5	20	1139	4885	32.85	18.71	0.77
5	30	1707	5224	44.94	26.25	0.93
5	40	2277	5565	57.25	33.88	1.10
5	50	2846	5904	69.77	41.61	1.27
5	74	4210	6715	99.34	59.96	1.68

End-To-End Benchmark Results

This page is intended to collect the latest end-to-end benchmarks results produced by Hydra's Continuous Integration system from the latest master code.

Please take those results with a grain of salt as they are currently produced from very limited cloud VMs and not controlled hardware. Instead of focusing on the absolute results, the emphasis should be on relative results, eg. how the timings for a scenario evolve as the code changes.

Generated at 2023-12-08 21:04:41.934947077 UTC

Baseline Scenario

Number of nodes	3
Number of txs	9000
Avg. Confirmation Time (ms)	23.447009153
P99	115.43640142000001ms
P95	33.225534399999994ms
P50	20.812751ms
Number of Invalid txs	0

Baseline Scenario

Number of nodes	1
Number of txs	3000
Avg. Confirmation Time (ms)	4.453277484
P99	7.309544409999993ms
P95	5.25924225ms
P50	4.339461999999999ms
Number of Invalid txs	0

ch1bo

There are still some things to clean up. All in all, the change looks better than expected. I have left several suggestions, but plan to address them also myself.

There is also quite some refactoring potential and i will try to remove the Hydra.Chain.Direct.State module completely from the "production code". That is, only use it to generate transactions in tx-cost and the hydra-node test suite.

hydra-node/src/Hydra/Chain/Direct.hs

hydra-node/src/Hydra/Chain/Direct/Tx.hs

ch1bo · 2023-12-05T09:48:43Z

hydra-node/src/Hydra/HeadLogic.hs

  notifyClient = ClientEffect $ ServerOutput.Committed headId pt utxo

  postCollectCom =
    OnChainEffect
-      { postChainTx = CollectComTx $ fold newCommitted
+      { postChainTx = CollectComTx{headId, headParameters = parameters}


Odd that we had constructed the tx with expected commits before, but now we seemingly infer them from the spent UTxO. But on the other hand parameters were added here instead of inferring them from the spendable UTxO.

Also, the AbortTx does have the committed UTxO.. this is very inconsistent.

Must: be consistent in one or the other approach

hydra-cardano-api/src/Hydra/Cardano/Api/PolicyId.hs

hydra-cluster/test/Test/DirectChainSpec.hs

ch1bo · 2023-12-05T14:47:24Z

hydra-node/src/Hydra/Chain/Direct/Handlers.hs

@@ -189,12 +197,12 @@ finalizeTx ::
  MonadThrow m =>
  TinyWallet m ->
  ChainContext ->
-  ChainStateType Tx ->
+  UTxO.UTxO ->
  UTxO.UTxO ->


Should: document which UTxO is which in haddock (it's an impossible to understand signature otherwise)

Do we need to separate it even?

i think we want to separate them, so to favor documentation... but you are correct; we can combine them all and have the function to take a single argument.

wdyt?

hydra-node/test/Hydra/Chain/Direct/StateSpec.hs

hydra-node/src/Hydra/Chain/Direct/State.hs

ch1bo

I am happy with the changes now!

Ideally we would update the CHANGELOG and bump versions of packages accordingly.

Let's merge this!

github-actions · 2023-12-07T19:57:08Z

Test Results

377 tests +5 372 ✔️ +5 21m 2s ⏱️ -49s
128 suites +2     5 💤 ±0
    5 files ±0     0 ❌ ±0

Results for commit ead4dab. ± Comparison against base commit 06feee9.

This pull request removes 9 and adds 14 tests. Note that renamed tests count towards both.

Hydra.Chain.Direct.Handlers/chainSyncHanlder ‑ observes transactions onRollForward
Hydra.Chain.Direct.Handlers/chainSyncHanlder ‑ roll forward fails with outdated TimeHandle
Hydra.Chain.Direct.Handlers/chainSyncHanlder ‑ roll forward results in Tick events
Hydra.Chain.Direct.Handlers/chainSyncHanlder ‑ rollbacks state onRollBackward
Hydra.Chain.Direct.State/abort ‑ ignore aborts of other heads
Hydra.Chain.Direct.State/init ‑ is not observed if not invited
Hydra.Chain.Direct.Tx/collectComTx ‑ Ignore InitTx with wrong cardano keys
Hydra.Chain.Direct.Tx/collectComTx ‑ Ignore InitTx with wrong contestation period
Hydra.HeadLogic/Coordinated Head Protocol ‑ ignores closeTx for another head

Hydra.Behavior/Two participant Head ‑ ignores head initialization of other head
Hydra.Chain.Direct.Handlers/chainSyncHandler ‑ observes transactions onRollForward
Hydra.Chain.Direct.Handlers/chainSyncHandler ‑ roll forward fails with outdated TimeHandle
Hydra.Chain.Direct.Handlers/chainSyncHandler ‑ roll forward results in Tick events
Hydra.Chain.Direct.Handlers/chainSyncHandler ‑ rollbacks state onRollBackward
Hydra.Chain.Direct.State/abort ‑ can create valid abort transactions for any observed head
Hydra.Chain.Direct.Tx/HeadId (cardano) ‑ headIdToPolicyId . mkHeadId === id
Hydra.Chain.Direct.Tx/HeadSeed (cardano) ‑ headSeedToTxIn . txInToHeadSeed === id
Hydra.HeadLogic/Coordinated Head Protocol ‑ ignores abortTx of another head
Hydra.HeadLogic/Coordinated Head Protocol ‑ ignores closeTx of another head
…

♻️ This comment has been updated with latest results.

This is mocking out the usage sites of the chain state on the transaction creation site.

As we removed those guards from the chain layer, it is important that the logic layer is doing this filtering.

…dLogic update

…Logic update

We were missing to check the headId from the observation with ours within the local state.

…reetings message'

In order to avoid breaking changes in the Greetings output messages introduce a new projection. This projection only gives you a HeadId if the Head is in Initializing state.

This is needed to ignore OnInitTx because it was not using the right on-chain keys.

We were only using observeRawInitTx and all checking of whether "its the right head" needs to be in the logic layer now. This commit does not yet migrate the test for this (which was in TxSpec)

This was done by the chain layer before, but needs to be done now by the logic layer. This is the only usage of on-chain identities in the off-chain Head protocol, but it still is required to retain semantics from before. This made initializing the Environment a bit hacky, as it is very specific to the Cardano configuration.

Originally only needed for the observation side (OnInitTx), the BehaviorSpec actually nudges us into also specifying the 'participants' on the creation side (InitTx). This feels right, despite it requiring quite a lot of changes in our test suites, because we can remove the other cardano keys from the direct chain handle! At some point this will enable us to open a head on-demand with any other participants.

This avoids a breaking change on the API, but keeps additional information about parties and contestationPeriod. This is helpful as all three things need to match.

We cannot use "derived" on chain identifiers in the Model tests, because we are actually validating the transactions and need to sign them correctly. Hence, the participation tokens need to be match the verification keys used.

We must reset the headId when the head is open or aborted, but keep it on all other server outputs.

This was correct originally, but got re-added incorrectly a few commits back

Avoid some diff to master and fix comments

While the DirectChainSpec tests would not need it, the EndToEnd scenarios have a fulll hydra-node running where the chain context and wallet still do load these files.

This was the last step to make all e2e tests green (two heads on the same network turned out to be flaky).

v0d1ch force-pushed the ensemble/stateless-observation-hydra-node branch from fb1969d to 1bda91b Compare December 5, 2023 09:26

ch1bo mentioned this pull request Dec 5, 2023

Stateless chain observation + tool #1096

Closed

3 tasks

ch1bo linked an issue Dec 5, 2023 that may be closed by this pull request

Stateless chain observation + tool #1096

Closed

3 tasks

ch1bo self-requested a review December 5, 2023 10:37

ch1bo assigned ch1bo, v0d1ch and ffakenz Dec 5, 2023

v0d1ch force-pushed the ensemble/stateless-observation-hydra-node branch from b448dca to e2c4e99 Compare December 5, 2023 15:00

ch1bo reviewed Dec 5, 2023

View reviewed changes

ffakenz force-pushed the ensemble/stateless-observation-hydra-node branch from b9438f8 to 715013d Compare December 5, 2023 18:26

ch1bo force-pushed the ensemble/stateless-observation-hydra-node branch from c5f3319 to 2399328 Compare December 6, 2023 07:27

v0d1ch force-pushed the ensemble/stateless-observation-hydra-node branch from 2316d04 to d99092c Compare December 6, 2023 09:34

ch1bo force-pushed the ensemble/stateless-observation-hydra-node branch from af7bbc4 to 232a5ad Compare December 6, 2023 17:09

ffakenz force-pushed the ensemble/stateless-observation-hydra-node branch 3 times, most recently from 251d97b to 1fd440c Compare December 6, 2023 19:34

ch1bo force-pushed the ensemble/stateless-observation-hydra-node branch from 5f9faa1 to 4713778 Compare December 7, 2023 14:04

v0d1ch force-pushed the ensemble/stateless-observation-hydra-node branch 2 times, most recently from 8788952 to b12a5a3 Compare December 7, 2023 17:21

ch1bo force-pushed the ensemble/stateless-observation-hydra-node branch 3 times, most recently from 002957b to 7299a5e Compare December 7, 2023 18:51

ch1bo approved these changes Dec 7, 2023

View reviewed changes

v0d1ch force-pushed the ensemble/stateless-observation-hydra-node branch from bf83efd to ce00931 Compare December 8, 2023 09:07

v0d1ch and others added 4 commits December 8, 2023 18:27

Use observeHeadTx instead of observeSomeTx in Chain.Direct.Handlers

4ab184a

Implement a convertObservation to match OnHeadTx interface

d89e0fd

Switch ChainStateAt to just hold a UTxO

0e72daf

This is mocking out the usage sites of the chain state on the transaction creation site.

Mock out ChainState usage in prepareTxToPost

5f372ae

ch1bo and others added 27 commits December 8, 2023 18:36

Add HeadLogic tests about ignoring observations of other heads

258594d

As we removed those guards from the chain layer, it is important that the logic layer is doing this filtering.

Add headId to OnContestTx and validate against ClosedState during Hea…

0e4cc50

…dLogic update

Add headId to OnFanoutTx and validate against ClosedState during Head…

41f152e

…Logic update

Fix headlogic for collect com

a175e14

We were missing to check the headId from the observation with ours within the local state.

Revert "Fix the Greetings api schema"

ed4f0c1

Revert 'Use the new HeadStatus scheme to fix the deserialization of G…

b356a5f

…reetings message'

Introduce HeadId Projection and use it

d2667f0

In order to avoid breaking changes in the Greetings output messages introduce a new projection. This projection only gives you a HeadId if the Head is in Initializing state.

Re-instantiate an OnChainId

d8bbf22

This is needed to ignore OnInitTx because it was not using the right on-chain keys.

Remove "non-raw" observeInitTx

f4673b8

We were only using observeRawInitTx and all checking of whether "its the right head" needs to be in the logic layer now. This commit does not yet migrate the test for this (which was in TxSpec)

Re-add participants to IgnoredHeadInitializing

69d080b

This avoids a breaking change on the API, but keeps additional information about parties and contestationPeriod. This is helpful as all three things need to match.

Add utxo to CollectComTx

91f4dd3

Fix ModelSpec by using concrete OnChainId

8ad2121

We cannot use "derived" on chain identifiers in the Model tests, because we are actually validating the transactions and need to sign them correctly. Hence, the participation tokens need to be match the verification keys used.

Start with fixing DirectChainSpec

89c345a

Add Carol as participant in a DirectChainSpec test

3df1b7c

Match participants and parties even when not in same order

38683b4

Fix and rename projectInitializingHeadId

3d5b2c9

We must reset the headId when the head is open or aborted, but keep it on all other server outputs.

Fix CollectCom logic

812f3e4

This was correct originally, but got re-added incorrectly a few commits back

Resolve warnings

95936f8

Formatting and minor fixes

657ccd4

Avoid some diff to master and fix comments

Still write cardano verification keys to disk in chainConfigFor

f88d440

While the DirectChainSpec tests would not need it, the EndToEnd scenarios have a fulll hydra-node running where the chain context and wallet still do load these files.

Add OnChainId to the api schema

12eaff7

Bump affected package versions

ef31640

Add changelog entry

c90b901

Add headId to OnChainTx

8f5f024

This was the last step to make all e2e tests green (two heads on the same network turned out to be flaky).

Fixes after rebasing

ead4dab

v0d1ch force-pushed the ensemble/stateless-observation-hydra-node branch from 6b5f8c4 to ead4dab Compare December 8, 2023 19:08

v0d1ch merged commit 6e7b43a into master Dec 8, 2023
21 of 22 checks passed

v0d1ch deleted the ensemble/stateless-observation-hydra-node branch December 8, 2023 21:27

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Stateless observation hydra node #1196

Stateless observation hydra node #1196

v0d1ch commented Dec 5, 2023 •

edited

Loading

github-actions bot commented Dec 5, 2023 •

edited

Loading

ch1bo left a comment

ch1bo Dec 5, 2023

ch1bo Dec 5, 2023

ch1bo Dec 5, 2023

ffakenz Dec 5, 2023

ch1bo left a comment

github-actions bot commented Dec 7, 2023 •

edited

Loading

Stateless observation hydra node #1196

Stateless observation hydra node #1196

Conversation

v0d1ch commented Dec 5, 2023 • edited Loading

github-actions bot commented Dec 5, 2023 • edited Loading

Transactions Costs

Script summary

Cost of Init Transaction

Cost of Commit Transaction

Cost of CollectCom Transaction

Cost of Close Transaction

Cost of Contest Transaction

Cost of Abort Transaction

Cost of FanOut Transaction

End-To-End Benchmark Results

Baseline Scenario

Baseline Scenario

ch1bo left a comment

Choose a reason for hiding this comment

ch1bo Dec 5, 2023

Choose a reason for hiding this comment

ch1bo Dec 5, 2023

Choose a reason for hiding this comment

ch1bo Dec 5, 2023

Choose a reason for hiding this comment

ffakenz Dec 5, 2023

Choose a reason for hiding this comment

ch1bo left a comment

Choose a reason for hiding this comment

github-actions bot commented Dec 7, 2023 • edited Loading

Test Results

v0d1ch commented Dec 5, 2023 •

edited

Loading

github-actions bot commented Dec 5, 2023 •

edited

Loading

github-actions bot commented Dec 7, 2023 •

edited

Loading