List stake distribution and stake pool statistics #713

Anviking · 2019-09-13T17:21:07Z

Context

ℹ️ This ticket is now an aggregation of:

Jörmungandr provides an endpoint for this with the following yaml-ified response format:

epoch: 228      # Epoch of last block
stake:
  dangling: 0 # Total value stored in accounts, but assigned to nonexistent pools
  pools:
    - - 5cf03f333f37eb7b987dbc9017b8a928287a3d77d086cd93cd9ad05bcba7e60f # stake pool ID
      - 1000000000000                                                    # staked value
    - - 3815602c096fcbb91072f419c296c3dfe1f730e0f446a9bd2553145688e75615 # stake pool ID
      - 1000000000000                                                    # staked value
  unassigned: 0 # Total value stored in accounts, but not assigned to any pool

Todo: When run with BFT-config, jörmungandr only returns {"epoch":2521}. How do we handle that?

Decision

Extend NetworkLayer and JörmungandrLayer to allow fetching this endpoint.

It should have a return-type similar to

(EpochNumber, [(PoolId, Stake)])
where
newtype Stake = Stake { getStakeLovelace :: Natural }
newtype PoolId = PoolVRFPubKey ByteString -- 32 bytes long

Acceptance Criteria

Our Jörmungandr client must be able to fetch stake pool distribution
We must be able to retrieve stake of each pool and the epoch for which the results are valid
We should be able to ignore dangling and unassigned stake
Users must be able to list stake pools via the CLI
It must be possible to eventually get an up-to-date distribution and statistics even with the events of rollbacks
It must be possible to know how many blocks each stake pool in the current epoch has produced
Distribution and pool statistics must be persisted so that it can be restored swiftly between restarts.
Wallet API clients must be able to list stake-pools and get their id, approximate performance and stake.
If API results are unavailable, this must be presented appropriately to the user.
The API response should not be paginated
Swagger.yaml must be updated

Development Plan

PR

Number	Base
#?	`master`

QA

integration tests, checking that we can list stake-pools, some correctness of data, and for error message when results are unavailable https://github.com/input-output-hk/cardano-wallet/blob/8ba0c261e7ae3a687c218db457b4e079e63b8155/lib/jormungandr/test/integration/Test/Integration/Jormungandr/Scenario/API/StakePools.hs#L43-L82
There exist properties for the Sqlite DB -> https://github.com/input-output-hk/cardano-wallet/blob/8ba0c261e7ae3a687c218db457b4e079e63b8155/lib/core/test/unit/Cardano/Pool/DB/Properties.hs#L95-L128
Two interesting cases of combineMetrics is tested https://github.com/input-output-hk/cardano-wallet/blob/8ba0c261e7ae3a687c218db457b4e079e63b8155/lib/core/test/unit/Cardano/Pool/MetricsSpec.hs#L49-L72
There might not be be race-conditions anymore https://github.com/input-output-hk/cardano-wallet/blob/8ba0c261e7ae3a687c218db457b4e079e63b8155/lib/core/src/Cardano/Pool/Metrics.hs#L170-L173

The text was updated successfully, but these errors were encountered:

Anviking · 2019-09-16T08:42:21Z

I'm not sure why jcli rest v0 stake get -h http://127.0.0.1:8080/api doesn't work for me on jcli 0.3.3, but curl-ing works:

curl http://127.0.0.1:8080/api/v0/stake
{"epoch":0,"stake":{"dangling":0,"pools":[],"unassigned":100100000000000}}⏎

KtorZ · 2019-09-17T07:30:11Z

From @rvl

yamlification was a good idea
Put in the type signature of the function that will be implemented

Anviking · 2019-09-18T06:38:08Z

yamlification was a good idea

Not quite sure what this refers to 🤔, but I simply copied the yaml-example from the jörmungandr docs. I wasn't advocating that we should serve yaml (in #715 that would be).

Anviking · 2019-09-18T11:15:03Z

About what is included in the total stake: The U/S clearly specifies S as the total stake currently delegated to pools. So we shouldn't have to care about unassigned or dangling.

Running jörmungandr with the following stake distribution:

$ curl http://127.0.0.1:8080/api/v0/stake
{"epoch":252054,"stake":{"dangling":0,"pools":[["7d749ef424507fb80fed0d2289d535a94f6870add0cf8b374cfe6cae078320ec",1]],"unassigned":100100000000000}}⏎

it produces blocks every slot.

Also, if the single pool has 0 stake, it doesn't produce blocks.

KtorZ · 2019-09-18T12:19:27Z

Also, if the single pool has 0 stake, it doesn't produce blocks.

👍 sounds intuitive.

763: add jormungandr get stake distribution r=KtorZ a=paweljakubas # Issue Number  #713 # Overview  - [x] I have added GET handle of jormungandr `/stake` - [x] I have added tests # Comments   764: Update Buildable instances for BlockHeader and Hash r=KtorZ a=Anviking # Issue Number #711 # Overview - [x] I added an idea for new `BlockHeader` buildable instance where the prevBlockHash comes first, and the blockHeader points towards it. - [x] I made `Hash tag` `build` to `424c4f43` instead of `424c4f43...c0dd1877`. - [x] I implemented `To/FromText` and `Buildable` instances on `Hash tag` in general, instead of having one instance per tag. # Comments  ```haskell fmtLn $ build h -- Old: Block (5.36#1036) -> 424c4f43...434b3033 -- New: 424c4f43 ⬸ [5.36#1036] fmtLn $ build (prevBlockHash h) -- Old: 424c4f43...434b3033 -- New: 424c4f43 ```  770: Foundations for stake-pools integration tests r=KtorZ a=piotr-iohk # Issue Number #715 # Overview  - [ ] I have added simple _foundations_ for future integration tests for stake-pools list # Comments   Co-authored-by: Pawel Jakubas <pawel.jakubas@iohk.io> Co-authored-by: Johannes Lund <johannes.lund@iohk.io> Co-authored-by: KtorZ <matthias.benkort@gmail.com> Co-authored-by: Piotr Stachyra <piotr.stachyra@iohk.io>

763: add jormungandr get stake distribution r=KtorZ a=paweljakubas # Issue Number  #713 # Overview  - [x] I have added GET handle of jormungandr `/stake` - [x] I have added tests # Comments   Co-authored-by: Pawel Jakubas <pawel.jakubas@iohk.io>

842: Add `NetworkLayer.stakeDistribution` and change types of response r=Anviking a=Anviking # Issue Number #713 # Overview  - [x] I simplified the types of JormungandrClient.getStakeDistribution and added NetworkLayer.stakeDistribution since our code currently works well with a large NetworkLayer. This way `StakePool.Metrics` can get the stake distribution without having a JörmungandrClient. - [ ] *Bonus*: <s>Add Buildable instance for Percentage</s> # Comments   Co-authored-by: Johannes Lund <johannes.lund@iohk.io>

856: implement 'withinSameTip' = to retry an action if the tip has changed while executing it r=Anviking a=KtorZ # Issue Number  #713 # Overview  - [x] I have added a helper function `withinSameTip` to "bracket" an action, retrying it if necessary (i.e. if the tip has changed during the execution of the action). # Comments  example usage: ```hs withinSameTip defaultPolicy (getNetworkTip nl) $ \tip -> do activity <- getStakePoolActivity tip distribution <- getStakeDistribution tip pure $ combineMetrics activity distribution ```  Co-authored-by: KtorZ <matthias.benkort@gmail.com>

856: implement 'withinSameTip' = to retry an action if the tip has changed while executing it r=KtorZ a=KtorZ # Issue Number  #713 # Overview  - [x] I have added a helper function `withinSameTip` to "bracket" an action, retrying it if necessary (i.e. if the tip has changed during the execution of the action). # Comments  example usage: ```hs withinSameTip defaultPolicy (getNetworkTip nl) $ \tip -> do activity <- getStakePoolActivity tip distribution <- getStakeDistribution tip pure $ combineMetrics activity distribution ```  Co-authored-by: KtorZ <matthias.benkort@gmail.com>

861: Measure intersection hit rate in chain following tests r=KtorZ a=KtorZ # Issue Number  #713 # Overview  - [x] I have added a `findIntersection` function to the network layer in order to allow client code to find an intersection with the current node's unstable block headers. - [x] I have measured the "intersection hit rate" (i.e., the number of time we are actually within `k` blocks from the tip before we call `nextBlocks` for the next batch). - [x] I have fixed a "small" bug in the http-bridge init cursor which would cause a partial pattern match to be triggered when the wallet is restarted. This gives us some idea of how well our chain following code performs. The main assumption is that our Jörmungandr client is actually much harsher than the reality. Therefore, if our chain follower does good in these scenarios, we might expect it to do rather okay in reality # Comments  We were wondering how frequent would we return a "ErrNoSynced" when fetching the stake pool distribution. We return such error if the our local chain has no intersection with the node's. With a small (in front of `k`) batch size, we currently get something like this: ``` 57.425% Intersection hit rate GREAT (75% - 100%) 32.235% Intersection hit rate GOOD (50% - 75%) 10.331% Intersection hit rate POOR (10% - 50%) 0.009% Intersection hit rate BAD (0% - 10%) ``` Now, if we increase the batch size to be bigger than k (like an order of magnitude) ``` 84.438% Intersection hit rate GREAT (75% - 100%) 15.541% Intersection hit rate GOOD (50% - 75%) 0.021% Intersection hit rate POOR (10% - 50%) ```  Co-authored-by: KtorZ <matthias.benkort@gmail.com>

868: store 'StakeDistribution' in the database, implement put & read ops r=KtorZ a=KtorZ # Issue Number  #713 # Overview  - [x] I have added a StakeDistribution table - [x] I have added two operations for reading and putting stake distributions to DB - [x] I have added two basic properties for the operations above. # Comments  ``` Sqlite Stake Pool properties readStakeDistribution . putStakeDistribution == pure +++ OK, passed 100 tests (4% Empty distributions). readStake . putStake a1 . putStake s0 == pure a1 +++ OK, passed 100 tests: 6% b is empty 4% a is empty 3% a & b are empty ```  Co-authored-by: KtorZ <matthias.benkort@gmail.com>

881: Connect Pool DB with Api r=KtorZ a=Anviking # Issue Number #713 # Overview - [x] I have implemented a `GET /stake-pool` that reads and combines metrics from `Pool.DB` - [x] TODO: Figure out when we are unsynced - [ ] TODO: Look for things to test if possible in this PR # Comments ### Self-node with a very short epoch-length (results are unavailable across epoch boundaries) ![self-node-short-epochs](https://user-images.githubusercontent.com/304423/67398367-d0edc680-f5aa-11e9-86da-ac13337e2a63.gif) ### Testnet syncing which is interrupted and resumed ![testnet-2](https://user-images.githubusercontent.com/304423/67399072-f7f8c800-f5ab-11e9-9fb1-38e8886aeba4.gif)   Co-authored-by: Johannes Lund <johannes.lund@iohk.io> Co-authored-by: KtorZ <matthias.benkort@gmail.com>

881: Connect Pool DB with Api r=Anviking a=Anviking # Issue Number #713 # Overview - [x] I have implemented a `GET /stake-pool` that reads and combines metrics from `Pool.DB` - [x] TODO: Figure out when we are unsynced - [ ] TODO: Look for things to test if possible in this PR # Comments ### Self-node with a very short epoch-length (results are unavailable across epoch boundaries) ![self-node-short-epochs](https://user-images.githubusercontent.com/304423/67398367-d0edc680-f5aa-11e9-86da-ac13337e2a63.gif) ### Testnet syncing which is interrupted and resumed ![testnet-2](https://user-images.githubusercontent.com/304423/67399072-f7f8c800-f5ab-11e9-9fb1-38e8886aeba4.gif)   Co-authored-by: Johannes Lund <johannes.lund@iohk.io> Co-authored-by: KtorZ <matthias.benkort@gmail.com>

881: Connect Pool DB with Api r=KtorZ a=Anviking # Issue Number #713 # Overview - [x] I have implemented a `GET /stake-pool` that reads and combines metrics from `Pool.DB` - [x] TODO: Figure out when we are unsynced - [ ] TODO: Look for things to test if possible in this PR # Comments ### Self-node with a very short epoch-length (results are unavailable across epoch boundaries) ![self-node-short-epochs](https://user-images.githubusercontent.com/304423/67398367-d0edc680-f5aa-11e9-86da-ac13337e2a63.gif) ### Testnet syncing which is interrupted and resumed ![testnet-2](https://user-images.githubusercontent.com/304423/67399072-f7f8c800-f5ab-11e9-9fb1-38e8886aeba4.gif)   Co-authored-by: Johannes Lund <johannes.lund@iohk.io> Co-authored-by: KtorZ <matthias.benkort@gmail.com>

910: Review stake pool worker creation r=Anviking a=KtorZ # Issue Number  #713 # Overview  - [x] I have slightly re-organized the code around the declaration of the stake pool layer - [x] I have namespaced the stake pool worker - [x] I have mounted some simple error handling on the stake pool worker thread # Comments   Co-authored-by: KtorZ <matthias.benkort@gmail.com>

919: Test stake-pool metrics on extremely short epochs r=Anviking a=Anviking # Issue Number  #713 # Overview  - [x] ⚠️ I have changed the epoch length to 3, and slotDuration to 1, in the genesis block of the integration tests. - [x] I have extended the expectation that we `eventually get 200`, to `eventually get 200 >> eventually get 503`, as the stake-pool metrics become unsynced when the node enters a new epoch. - [x] I have added guarantees on the returned list of stake-pools based on our specific testing setup. # Comments   Co-authored-by: Johannes Lund <johannes.lund@iohk.io>

928: Test that monitorStakePools resumes correctly after being killed r=Anviking a=Anviking # Issue Number #713 # Overview - [x] I added an integration test, testing `monitorStakePools` and `Pool.DB, making sure that if a worker is killed and restarted, it is the same as if it had never been killed. # Comments  Co-authored-by: Johannes Lund <johannes.lund@iohk.io>

940: Add more stake-pools in the CI setup r=Anviking a=Anviking # Issue Number #713 # Overview - [x] I have added more stake-pools to the CI setup. We launch 1 pool, and have 2 inactive ones. # Comments - I don't really understand the effect on `consensus_genesis_praos_active_slot_coeff` on the block-production / slot, I'm checking with the rust folks. - As of currently this PR keeps `consensus_genesis_praos_active_slot_coeff : 1` and results in 100% of the slots being produced by the first pool, despite having 33% of the stake.   Co-authored-by: Johannes Lund <johannes.lund@iohk.io>

924: Prototype implementation of approximate pool performance r=KtorZ a=Anviking # Issue Number #713 # Overview  - [x] Very quick prototype implementation of approximate performance (`AP`). - [x] Performance is included in the stake-pool-metrics served by the api. # Comments  - Should make sure corner-cases are handled correctly (numbers may be 0) - On the jormungandr testnet one pool has a performance of 53, while the others have 0 🤔 . (There is only one pool that is actually producing blocks.)  Co-authored-by: Johannes Lund <johannes.lund@iohk.io> Co-authored-by: KtorZ <matthias.benkort@gmail.com>

979: Support atomic and isolated executions of the database layers r=KtorZ a=Anviking # Issue Number #713 # Overview - [x] I have made sure that `putPoolProduction` and `putStakeDistribution` happens in a single Sqlite transaction. If the wallet were to crash, either none of them or both of them succeed. - [x] Existing tests verify that this works for large batches `~ 100`. I tried writing 2000 pool-productions atomically, and it also worked, but slowed down tests, so I didn't commit it. # Comments   Co-authored-by: Johannes Lund <johannes.lund@iohk.io> Co-authored-by: KtorZ <matthias.benkort@gmail.com>

KtorZ · 2019-11-20T19:47:30Z

Shall we close that one 😶 ?

piotr-iohk · 2019-11-21T10:02:14Z

Yes I think so, I was just keeping it here while doing some exploratory testing on different testnets (which discovered few issues)...

Anviking · 2019-11-21T12:44:28Z

It might be worth attempting to generalise the performance calculation (to e.g. support exponential moving averages) and at the same time extending the tests.

KtorZ · 2019-11-21T14:27:29Z

to generalise the performance calculation

To cover what business requirement exactly 🤔 ?

at the same time extending the tests.

I think we already had 50 days for that :) ... Any more work on this would go in a new story, in a new ticket.

KtorZ added this to the List Stake Pool milestone Sep 13, 2019

Anviking mentioned this issue Sep 15, 2019

Calculate performance and list stake pools in the WalletLayer #714

Closed

4 tasks

Anviking changed the title ~~Get stake-distribution from Jörmungandr's REST API~~ Get stake-distribution from Jörmungandr's GET /stake Sep 15, 2019

Anviking changed the title ~~Get stake-distribution from Jörmungandr's GET /stake~~ Get stake-distribution from Jörmungandr's GET /stake Sep 18, 2019

Anviking mentioned this issue Sep 18, 2019

Count how many blocks each pool produced — I. Simple approach #711

Closed

6 tasks

Anviking mentioned this issue Sep 23, 2019

Add API endpoint for listing stake pools #715

Closed

4 tasks

paweljakubas mentioned this issue Sep 23, 2019

add api types and handlers for listStakePools endpoint #748

Merged

2 tasks

paweljakubas self-assigned this Sep 25, 2019

paweljakubas mentioned this issue Sep 27, 2019

add jormungandr get stake distribution #763

Merged

2 tasks

Anviking mentioned this issue Oct 16, 2019

Add NetworkLayer.stakeDistribution and change types of response #842

Merged

2 tasks

KtorZ mentioned this issue Oct 17, 2019

implement 'withinSameTip' = to retry an action if the tip has changed while executing it #856

Merged

1 task

KtorZ mentioned this issue Oct 17, 2019

Measure intersection hit rate in chain following tests #861

Merged

3 tasks

KtorZ mentioned this issue Oct 18, 2019

store 'StakeDistribution' in the database, implement put & read ops #868

Merged

3 tasks

KtorZ mentioned this issue Oct 21, 2019

Count how many blocks each pool produced — III. Sqlite #712

Closed

4 tasks

KtorZ mentioned this issue Oct 24, 2019

Review stake pool worker creation #910

Merged

3 tasks

This was referenced Oct 25, 2019

Test stake-pool metrics on extremely short epochs #919

Merged

Prototype implementation of approximate pool performance #924

Merged

Anviking mentioned this issue Oct 28, 2019

Test that monitorStakePools resumes correctly after being killed #928

Merged

1 task

Anviking mentioned this issue Oct 30, 2019

Add more stake-pools in the CI setup #940

Merged

1 task

Anviking mentioned this issue Nov 7, 2019

Support atomic and isolated executions of the database layers #979

Merged

2 tasks

piotr-iohk closed this as completed Nov 21, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

List stake distribution and stake pool statistics #713

List stake distribution and stake pool statistics #713

Anviking commented Sep 13, 2019 •

edited

Loading

Anviking commented Sep 16, 2019

KtorZ commented Sep 17, 2019

Anviking commented Sep 18, 2019

Anviking commented Sep 18, 2019

KtorZ commented Sep 18, 2019

KtorZ commented Nov 20, 2019

piotr-iohk commented Nov 21, 2019

Anviking commented Nov 21, 2019

KtorZ commented Nov 21, 2019

List stake distribution and stake pool statistics #713

List stake distribution and stake pool statistics #713

Comments

Anviking commented Sep 13, 2019 • edited Loading

Context

Decision

Acceptance Criteria

Development Plan

PR

QA

Anviking commented Sep 16, 2019

KtorZ commented Sep 17, 2019

Anviking commented Sep 18, 2019

Anviking commented Sep 18, 2019

KtorZ commented Sep 18, 2019

KtorZ commented Nov 20, 2019

piotr-iohk commented Nov 21, 2019

Anviking commented Nov 21, 2019

KtorZ commented Nov 21, 2019

Anviking commented Sep 13, 2019 •

edited

Loading