Account for ouroboros-network changes

* Update `ouroboros-consensus-diffusion` to `ouroboros-network-0.16.0.0` * GSM: use diffusion layer info for HAA Concretely, this addresses #974 in the context of bootstrap peers, but this code will also work for the "actual" Honest Availability Assumption of Genesis once implemented on the Network side. * BootstrapPeers.md: include pseudo-HAA * Remove `ShowProxy` instance of `SlotNo` It is now provided by `ouroboros-network`.
IntersectMBO · May 9, 2024 · 3f873d6 · 3f873d6
1 parent 4faa552
commit 3f873d6
Show file tree

Hide file tree

Showing 9 changed files with 82 additions and 66 deletions.
diff --git a/cabal.project b/cabal.project
@@ -58,6 +58,16 @@ source-repository-package
     eras/byron/crypto
     eras/byron/crypto/test
 
+source-repository-package
+  type: git
+  location: https://github.com/IntersectMBO/ouroboros-network
+  tag: 821feaa29b9f82d1364345348b304484bc189283
+  --sha256: sha256-yQEj74pnTVJQeeHNRr5ejW1KSKp89OFbNROtOdthG/o=
+  subdir: ouroboros-network
+          ouroboros-network-api
+          ouroboros-network-framework
+          ouroboros-network-protocols
+
 -- We want to always build the test-suites and benchmarks
 tests: true
 benchmarks: true

diff --git a/docs/website/contents/for-developers/BootstrapPeersIER.md b/docs/website/contents/for-developers/BootstrapPeersIER.md
@@ -12,32 +12,48 @@ The following state machine depicts the desired behavior of the node.
 
 ```mermaid
 graph
-    OnlyBootstrap[OnlyBootstrap]
-    CaughtUp[CaughtUp]
+    subgraph OnlyBootstrap
+        direction TB
+        PreSyncing[PreSyncing]
+        Syncing[Syncing]
+        PreSyncing -- "Honest Availability Assumption\nis satisfied" --> Syncing
+        Syncing -- "Honest Availability Assumption\nis no longer satisfied" --> PreSyncing
+    end
 
-    OnlyBootstrap -- "no peers claim to have\nsubsequent headers,\nand its selection is ≥\nthe best header" --> CaughtUp
-    CaughtUp -- "vol tip became older than X" --> OnlyBootstrap
+    CaughtUp[CaughtUp]
+    Syncing -- "no peers claim to have\nsubsequent headers,\nand its selection is ≥\nthe best header" --> CaughtUp
+    CaughtUp -- "vol tip became older than X" ----> PreSyncing
 
     StartUp[[Node start-up]]
     StartUp -- "node was most recently in CaughtUp\nand vol tip is younger than X" --> CaughtUp
-    StartUp -- "otherwise" --> OnlyBootstrap
+    StartUp -- "otherwise" --> PreSyncing
 ```
 
-- `OnlyBootstrap` state - All upstream peers must reside in a centralized set of trusted _bootstrap peers_.
+- `OnlyBootstrap` - All upstream peers must be trusted.
+
+   In the context of bootstrap peers, as all peers are trusted, the _Honest Availability Assumption_ is satisfied in the following cases:
+
+    - The node is configured to connect to bootstrap peers, and it has established a connection to a bootstrap peer.
+
+    - The node is not configured to connect to bootstrap peers. This is the case for eg block producers and hidden relays. They will only be connected to trusted local root peers (eg the relays for a block-producing node).
 
 - `CaughtUp` state - The peers are chosen according to the P2P design, including the _ledger peers_ etc.
 
 **Desideratum 2.**
 In particular, the transitions should happen promptly.
 
-- `CaughtUp -> OnlyBootstrap` should be prompt in order to minimize the duration that the node is exposed to untrusted peers (aka non-bootstrap peers) while its stale volatile tip is making it vulnerable.
+- `CaughtUp -> PreSyncing` should be prompt in order to minimize the duration that the node is exposed to untrusted peers (aka non-bootstrap peers) while its stale volatile tip is making it vulnerable.
   Delays here would directly threaten the security of the node.
 
-- `OnlyBootstrap -> CaughtUp` should be prompt so that the centralized, relatively-few bootstrap peers are relieved of load as soon as possible.
+- `Syncing -> CaughtUp` should be prompt so that the centralized, relatively-few bootstrap peers are relieved of load as soon as possible.
   Delays here would not directly threaten the security of the node.
   However, wasting the centralized resources would threaten the ability of nodes to join the net, ie the availability of the whole net.
   Determining the exact load constraints for the bootstrap peers is not yet finalized.
 
+- `PreSyncing -> Syncing` should be prompt to allow the node to conclude that is is caught up as a follow-up.
+
+- `Syncing -> PreSyncing` should be prompt to prevent the node from concluding that it is caught up while it is not actually connected to a bootstrap peers.
+
 **Desideratum 3.**
 The node should not return to `OnlyBootstrap` every time it restarts/briefly loses network/etc.
 Such unnecessary connections would also put unnecessary load on the centralized, relatively-few bootstrap peers.
@@ -59,11 +75,11 @@ This is the point of the "Node start-up" pseudo state in the diagram above.
 
   As the volatile tip age approaches X, the Consensus Layer could forewarn the Diffusion Layer, eg "it seems like the transition back to OnlyBootstrap will be necessary soon; please prepare", if that would be helpful.
 
-- For similar reasons, the Diffusion Layer should also manage the disconections from all peers upon the `OnlyBootstrap -> CaughtUp` transition.
+- For similar reasons, the Diffusion Layer should also manage the disconections from all (bootstrap) peers upon the `OnlyBootstrap -> CaughtUp` transition.
 
 ## Anticipated Interface
 
-See [IntersectMBO/ouroboros-network#4555](https://github.com/IntersectMBO/ouroboros-network/pull/4555) for the definition/implementation of this interface on the Network side.
+See [IntersectMBO/ouroboros-network#4555](https://github.com/IntersectMBO/ouroboros-network/pull/4555) and [IntersectMBO/ouroboros-network#4846](https://github.com/IntersectMBO/ouroboros-network/pull/4846) for the definition/implementation of this interface on the Network side.
 
 - The Diffusion Layer should monitor a `TVar State` (maybe via a `STM State` action).
   The Consensus Layer will update that state promptly.
@@ -75,6 +91,12 @@ See [IntersectMBO/ouroboros-network#4555](https://github.com/IntersectMBO/ourobo
   Here, `YoungEnough` signals that the ledger state's distribution among stake relays is sufficiently close to that of the actual real world.
   For now, we conservatively return `YoungEnough` only when the node concludes it has fully caught-up, and `TooOld` otherwise.
 
+- The Diffusion Layer will inform the Consensus Layer whether the Honest Availability Assumption is satisfied.
+  ```haskell
+  data OutboundConnectionsState = TrustedStateWithExternalPeers | UntrustedState
+  daUpdateOutboundConnectionsState :: OutboundConnectionsState -> STM m ()
+  ```
+
 - Whenever necessary, the Diffusion Layer can ask the Consensus Layer for the ledger peer information, eg
   ```haskell
   lpGetLedgerPeers :: STM m [(PoolStake, NonEmpty RelayAccessPoint)]

diff --git a/...s-consensus-diffusion/changelog.d/20240429_143457_alexander.esgen_gsm_remote.md b/...s-consensus-diffusion/changelog.d/20240429_143457_alexander.esgen_gsm_remote.md
@@ -0,0 +1,5 @@
+### Non-Breaking
+
+- Implemented the Honest Availability Assumption properly (both for
+  Praos/"Genesis Lite" and Genesis) based on newly exposed state by the
+  diffusion layer.
diff --git a/...-consensus-diffusion/changelog.d/20240508_072753_coot_ouroboros_network_0_16.md b/...-consensus-diffusion/changelog.d/20240508_072753_coot_ouroboros_network_0_16.md
@@ -0,0 +1,3 @@
+### Non-Breaking
+
+- Upgraded to `ouroboros-network-0.16`
diff --git a/ouroboros-consensus-diffusion/ouroboros-consensus-diffusion.cabal b/ouroboros-consensus-diffusion/ouroboros-consensus-diffusion.cabal
@@ -89,10 +89,10 @@ library
     io-classes ^>=1.4.1,
     mtl,
     ouroboros-consensus ^>=0.17,
-    ouroboros-network ^>=0.14,
-    ouroboros-network-api ^>=0.7,
-    ouroboros-network-framework ^>=0.12,
-    ouroboros-network-protocols ^>=0.8,
+    ouroboros-network ^>=0.16,
+    ouroboros-network-api ^>=0.7.2,
+    ouroboros-network-framework ^>=0.13,
+    ouroboros-network-protocols ^>=0.8.1,
     random,
     safe-wild-cards ^>=1.0,
     serialise ^>=0.2,

diff --git a/ouroboros-consensus-diffusion/src/ouroboros-consensus-diffusion/Ouroboros/Consensus/Node.hs b/ouroboros-consensus-diffusion/src/ouroboros-consensus-diffusion/Ouroboros/Consensus/Node.hs
@@ -22,7 +22,6 @@ module Ouroboros.Consensus.Node (
     -- * Standard arguments
   , StdRunNodeArgs (..)
   , stdBfcSaltIO
-  , stdChainSyncTimeout
   , stdGsmAntiThunderingHerdIO
   , stdKeepAliveRngIO
   , stdLowLevelRunNodeArgsIO
@@ -60,6 +59,7 @@ import qualified Codec.CBOR.Encoding as CBOR
 import           Codec.Serialise (DeserialiseFailure)
 import qualified Control.Concurrent.Class.MonadSTM.Strict as StrictSTM
 import           Control.DeepSeq (NFData)
+import           Control.Monad (when)
 import           Control.Monad.Class.MonadTime.SI (MonadTime)
 import           Control.Monad.Class.MonadTimer.SI (MonadTimer)
 import           Control.Tracer (Tracer, contramap, traceWith)
@@ -110,9 +110,9 @@ import           Ouroboros.Consensus.Util.ResourceRegistry
 import           Ouroboros.Consensus.Util.Time (secondsToNominalDiffTime)
 import           Ouroboros.Network.BlockFetch (BlockFetchConfiguration (..))
 import qualified Ouroboros.Network.Diffusion as Diffusion
+import qualified Ouroboros.Network.Diffusion.Configuration as Diffusion
 import qualified Ouroboros.Network.Diffusion.NonP2P as NonP2P
 import qualified Ouroboros.Network.Diffusion.P2P as P2P
-import qualified Ouroboros.Network.Diffusion.Policies as Diffusion
 import           Ouroboros.Network.Magic
 import           Ouroboros.Network.NodeToClient (ConnectionId, LocalAddress,
                      LocalSocket, NodeToClientVersionData (..), combineVersions,
@@ -129,15 +129,14 @@ import           Ouroboros.Network.PeerSelection.PeerMetric (PeerMetrics,
 import           Ouroboros.Network.PeerSelection.PeerSharing (PeerSharing)
 import           Ouroboros.Network.PeerSelection.PeerSharing.Codec
                      (decodeRemoteAddress, encodeRemoteAddress)
-import           Ouroboros.Network.Protocol.Limits (shortWait)
 import           Ouroboros.Network.RethrowPolicy
 import qualified SafeWildCards
 import           System.Exit (ExitCode (..))
 import           System.FilePath ((</>))
 import           System.FS.API (SomeHasFS (..))
 import           System.FS.API.Types
 import           System.FS.IO (ioHasFS)
-import           System.Random (StdGen, newStdGen, randomIO, randomRIO, split)
+import           System.Random (StdGen, newStdGen, randomIO, split)
 
 {-------------------------------------------------------------------------------
   The arguments to the Consensus Layer node functionality
@@ -626,7 +625,11 @@ runWith RunNodeArgs{..} encAddrNtN decAddrNtN LowLevelRunNodeArgs{..} =
                   lpGetLatestSlot = getImmTipSlot kernel,
                   lpGetLedgerPeers = fromMaybe [] <$> getPeersFromCurrentLedger kernel (const True),
                   lpGetLedgerStateJudgement = getLedgerStateJudgement kernel
-                }
+                },
+            Diffusion.daUpdateOutboundConnectionsState =
+              let varOcs = getOutboundConnectionsState kernel in \newOcs -> do
+                oldOcs <- readTVar varOcs
+                when (newOcs /= oldOcs) $ writeTVar varOcs newOcs
           }
 
         localRethrowPolicy :: RethrowPolicy
@@ -732,7 +735,7 @@ mkNodeKernelArgs
       , blockFetchSize          = estimateBlockSize
       , mempoolCapacityOverride = NoMempoolCapacityBytesOverride
       , miniProtocolParameters  = defaultMiniProtocolParameters
-      , blockFetchConfiguration = defaultBlockFetchConfiguration
+      , blockFetchConfiguration = Diffusion.defaultBlockFetchConfiguration bfcSalt
       , gsmArgs = GsmNodeKernelArgs {
           gsmAntiThunderingHerd
         , gsmDurationUntilTooOld
@@ -744,15 +747,6 @@ mkNodeKernelArgs
       , peerSharingRng = psRng
       , publicPeerSelectionStateVar
       }
-  where
-    defaultBlockFetchConfiguration :: BlockFetchConfiguration
-    defaultBlockFetchConfiguration = BlockFetchConfiguration
-      { bfcMaxConcurrencyBulkSync = 1
-      , bfcMaxConcurrencyDeadline = 1
-      , bfcMaxRequestsInflight    = fromIntegral $ blockFetchPipeliningMax defaultMiniProtocolParameters
-      , bfcDecisionLoopInterval   = 0.01 -- 10ms
-      , bfcSalt
-      }
 
 -- | We allow the user running the node to customise the 'NodeKernelArgs'
 -- through 'llrnCustomiseNodeKernelArgs', but there are some limits to some
@@ -800,33 +794,6 @@ stdGsmAntiThunderingHerdIO = newStdGen
 stdKeepAliveRngIO :: IO StdGen
 stdKeepAliveRngIO = newStdGen
 
-stdChainSyncTimeout :: IO NTN.ChainSyncTimeout
-stdChainSyncTimeout = do
-    -- These values approximately correspond to false positive
-    -- thresholds for streaks of empty slots with 99% probability,
-    -- 99.9% probability up to 99.999% probability.
-    -- t = T_s [log (1-Y) / log (1-f)]
-    -- Y = [0.99, 0.999...]
-    -- T_s = slot length of 1s.
-    -- f = 0.05
-    -- The timeout is randomly picked per bearer to avoid all bearers
-    -- going down at the same time in case of a long streak of empty
-    -- slots.
-    -- To avoid global synchronosation the timeout is picked uniformly
-    -- from the interval 135 - 269, corresponds to the a 99.9% to
-    -- 99.9999% thresholds.
-    -- TODO: The timeout should be drawn at random everytime chainsync
-    --       enters the must reply state. A static per connection timeout
-    --       leads to selection preassure for connections with a large
-    --       timeout, see #4244.
-    mustReplyTimeout <- Just <$> realToFrac <$> randomRIO (135,269 :: Double)
-    return NTN.ChainSyncTimeout
-      { canAwaitTimeout  = shortWait
-      , intersectTimeout = shortWait
-      , mustReplyTimeout
-      , idleTimeout      = Just 3673
-      }
-
 stdVersionDataNTN :: NetworkMagic
                   -> DiffusionMode
                   -> PeerSharing
@@ -887,7 +854,7 @@ stdLowLevelRunNodeArgsIO RunNodeArgs{ rnProtocolInfo
     llrnKeepAliveRng          <- stdKeepAliveRngIO
     pure LowLevelRunNodeArgs
       { llrnBfcSalt
-      , llrnChainSyncTimeout = fromMaybe stdChainSyncTimeout srnChainSyncTimeout
+      , llrnChainSyncTimeout = fromMaybe Diffusion.defaultChainSyncTimeout srnChainSyncTimeout
       , llrnChainSyncLoPBucketConfig = ChainSyncLoPBucketDisabled
       , llrnCustomiseHardForkBlockchainTimeArgs = id
       , llrnGsmAntiThunderingHerd

diff --git a/...s-consensus-diffusion/src/ouroboros-consensus-diffusion/Ouroboros/Consensus/NodeKernel.hs b/...s-consensus-diffusion/src/ouroboros-consensus-diffusion/Ouroboros/Consensus/NodeKernel.hs
@@ -94,6 +94,8 @@ import           Ouroboros.Network.NodeToNode (ConnectionId,
 import           Ouroboros.Network.PeerSelection.Bootstrap (UseBootstrapPeers)
 import           Ouroboros.Network.PeerSelection.LedgerPeers.Type
                      (LedgerStateJudgement (..))
+import           Ouroboros.Network.PeerSelection.LocalRootPeers
+                     (OutboundConnectionsState (..))
 import           Ouroboros.Network.PeerSharing (PeerSharingAPI,
                      PeerSharingRegistry, newPeerSharingAPI,
                      newPeerSharingRegistry, ps_POLICY_PEER_SHARE_MAX_PEERS,
@@ -149,6 +151,9 @@ data NodeKernel m addrNTN addrNTC blk = NodeKernel {
     , setBlockForging        :: [BlockForging m blk] -> m ()
 
     , getPeerSharingAPI      :: PeerSharingAPI addrNTN StdGen m
+
+    , getOutboundConnectionsState
+                             :: StrictTVar m OutboundConnectionsState
     }
 
 -- | Arguments required when initializing a node
@@ -204,6 +209,8 @@ initNodeKernel args@NodeKernelArgs { registry, cfg, tracers
           , varLedgerJudgement
           } = st
 
+    varOutboundConnectionsState <- newTVarIO UntrustedState
+
     do  let GsmNodeKernelArgs {..} = gsmArgs
             gsmTracerArgs          =
               ( castTip . either AF.anchorToTip tipFromHeader . AF.head . fst
@@ -239,9 +246,12 @@ initNodeKernel args@NodeKernelArgs { registry, cfg, tracers
                     gsmMarkerFileView
               , GSM.writeGsmState = \x -> atomically $ do
                   writeTVar varLedgerJudgement $ GSM.gsmStateToLedgerJudgement x
-              , -- In the context of bootstrap peers, it is fine to always
-                -- return 'True' as all peers are trusted during syncing.
-                GSM.isHaaSatisfied            = pure True
+              , GSM.isHaaSatisfied            = do
+                  readTVar varOutboundConnectionsState <&> \case
+                    -- See the upstream Haddocks for the exact conditions under
+                    -- which the diffusion layer is in this state.
+                    TrustedStateWithExternalPeers -> True
+                    UntrustedState                -> False
               }
         judgment <- readTVarIO varLedgerJudgement
         void $ forkLinkedThread registry "NodeKernel.GSM" $ case judgment of
@@ -278,6 +288,8 @@ initNodeKernel args@NodeKernelArgs { registry, cfg, tracers
       , getTracers              = tracers
       , setBlockForging         = \a -> atomically . LazySTM.putTMVar blockForgingVar $! a
       , getPeerSharingAPI       = peerSharingAPI
+      , getOutboundConnectionsState
+                                = varOutboundConnectionsState
       }
   where
     blockForgingController :: InternalState m remotePeer localPeer blk

diff --git a/ouroboros-consensus/changelog.d/20240509_022039_Lucsanszky_integration_8_11.md b/ouroboros-consensus/changelog.d/20240509_022039_Lucsanszky_integration_8_11.md
@@ -0,0 +1,3 @@
+### Breaking
+
+- Remove `ShowProxy` instance of `SlotNo` (now provided by `ouroboros-network`)
diff --git a/ouroboros-consensus/src/ouroboros-consensus/Ouroboros/Consensus/Util/Orphans.hs b/ouroboros-consensus/src/ouroboros-consensus/Ouroboros/Consensus/Util/Orphans.hs
@@ -64,12 +64,6 @@ instance Serialise (VerKeyDSIGN MockDSIGN) where
   encode = encodeVerKeyDSIGN
   decode = decodeVerKeyDSIGN
 
-{-------------------------------------------------------------------------------
-  ShowProxy
--------------------------------------------------------------------------------}
-
-instance ShowProxy SlotNo where
-
 {-------------------------------------------------------------------------------
   NoThunks
 -------------------------------------------------------------------------------}