Update aristo journal functionality (#2155)

* Aristo: Code cosmetics, e.g. update some CamelCase names

* CoreDb+Aristo: Provide oldest known state root implied

details:
  The Aristo journal allows to recover earlier but not all state roots.

* Aristo: Fix journal backward index operator, e.g. `[^1]`

* Aristo: Fix journal updater

why:
  The `fifosStore()` store function slightly misinterpreted the update
  instructions when translation is to database `put()` functions. The
  effect was that the journal was ever growing due to stale entries which
  were never deleted.

* CoreDb+Aristo: Provide utils for purging stale data from the KVT

details:
  See earlier patch, not all state roots are available. This patch
  provides a mapping from some state root to a block number and allows to
  remove all KVT data related to a particular block number

* Aristo+Kvt: Implement a clean up schedule for expired data in KVT

why:
  For a single state ledger like `Aristo`, there is only a limited
  backlog of states. So KVT data (i.e. headers etc.) are cleaned up
  regularly

* Fix copyright year

nimbus/core/chain/persist_blocks.nim

@@ -36,6 +36,10 @@ type
 
   PersistBlockFlags = set[PersistBlockFlag]
 
+const
+  CleanUpEpoch = 30_000.u256
+    ## Regular checks for history clean up (applies to single state DB)
+
 # ------------------------------------------------------------------------------
 # Private
 # ------------------------------------------------------------------------------
@@ -53,12 +57,22 @@ proc getVmState(c: ChainRef, header: BlockHeader):
     return err()
   return ok(vmState)
 
+proc purgeExpiredBlocks(db: CoreDbRef) {.inline, raises: [RlpError].} =
+  ## Remove non-reachable blocks from KVT database
+  var blkNum = db.getOldestJournalBlockNumber()
+  if 0 < blkNum:
+    blkNum = blkNum - 1
+    while 0 < blkNum:
+      if not db.forgetHistory blkNum:
+        break
+      blkNum = blkNum - 1
+
+
 proc persistBlocksImpl(c: ChainRef; headers: openArray[BlockHeader];
                        bodies: openArray[BlockBody],
                        flags: PersistBlockFlags = {}): ValidationResult
                           # wildcard exception, wrapped below in public section
                           {.inline, raises: [CatchableError].} =
-
   let dbTx = c.db.beginTransaction()
   defer: dbTx.dispose()
 
@@ -71,10 +85,13 @@ proc persistBlocksImpl(c: ChainRef; headers: openArray[BlockHeader];
   let vmState = c.getVmState(headers[0]).valueOr:
     return ValidationResult.Error
 
-  trace "Persisting blocks",
+  # Check point
-    fromBlock = headers[0].blockNumber,
+  let stateRootChpt = vmState.parent.stateRoot
-    toBlock = headers[^1].blockNumber
 
+  # Needed for figuring out whether KVT cleanup is due (see at the end)
+  let (fromBlock, toBlock) = (headers[0].blockNumber, headers[^1].blockNumber)
+
+  trace "Persisting blocks", fromBlock, toBlock
   for i in 0 ..< headers.len:
     let (header, body) = (headers[i], bodies[i])
 
@@ -177,6 +194,13 @@ proc persistBlocksImpl(c: ChainRef; headers: openArray[BlockHeader];
   # The `c.db.persistent()` call is ignored by the legacy DB which
   # automatically saves persistently when reaching the zero level transaction
   c.db.persistent()
+
+  # For a single state ledger, there is only a limited backlog. So clean up
+  # regularly (the `CleanUpEpoch` should not be too small as each lookup pulls
+  # a journal entry from disk.)
+  if (fromBlock mod CleanUpEpoch) <= (toBlock - fromBlock):
+    c.db.purgeExpiredBlocks()
+
   ValidationResult.OK
 
 # ------------------------------------------------------------------------------

nimbus/db/aristo/aristo_api.nim

@@ -148,6 +148,13 @@ type
       ## pair was found on the filter or the backend, this transaction is
       ## empty.
 
+  AristoApiGetFilUbeFn* =
+    proc(db: AristoDbRef;
+         qid: QueueID;
+        ): Result[FilterRef,AristoError]
+        {.noRaise.}
+      ## Get the filter from the unfiltered backened if available.
+
   AristoApiGetKeyRcFn* =
     proc(db: AristoDbRef;
          vid: VertexID;
@@ -352,6 +359,7 @@ type
     forget*: AristoApiForgetFn
     forkTop*: AristoApiForkTopFn
     forkWith*: AristoApiForkWithFn
+    getFilUbe*: AristoApiGetFilUbeFn
     getKeyRc*: AristoApiGetKeyRcFn
     hashify*: AristoApiHashifyFn
     hasPath*: AristoApiHasPathFn
@@ -384,6 +392,7 @@ type
     AristoApiProfForgetFn       = "forget"
     AristoApiProfForkTopFn      = "forkTop"
     AristoApiProfForkWithFn     = "forkWith"
+    AristoApiProfGetFilUbeFn    = "getFilUBE"
     AristoApiProfGetKeyRcFn     = "getKeyRc"
     AristoApiProfHashifyFn      = "hashify"
     AristoApiProfHasPathFn      = "hasPath"
@@ -434,6 +443,7 @@ when AutoValidateApiHooks:
     doAssert not api.forget.isNil
     doAssert not api.forkTop.isNil
     doAssert not api.forkWith.isNil
+    doAssert not api.getFilUbe.isNil
     doAssert not api.getKeyRc.isNil
     doAssert not api.hashify.isNil
     doAssert not api.hasPath.isNil
@@ -486,6 +496,7 @@ func init*(api: var AristoApiObj) =
   api.forget = forget
   api.forkTop = forkTop
   api.forkWith = forkWith
+  api.getFilUbe = getFilUbe
   api.getKeyRc = getKeyRc
   api.hashify = hashify
   api.hasPath = hasPath
@@ -521,6 +532,7 @@ func dup*(api: AristoApiRef): AristoApiRef =
     forget:       api.forget,
     forkTop:      api.forkTop,
     forkWith:     api.forkWith,
+    getFilUbe:    api.getFilUbe,
     getKeyRc:     api.getKeyRc,
     hashify:      api.hashify,
     hasPath:      api.hasPath,
@@ -609,6 +621,11 @@ func init*(
       AristoApiProfForkWithFn.profileRunner:
         result = api.forkWith(a, b, c, d)
 
+  profApi.getFilUbe =
+    proc(a: AristoDbRef; b: QueueID): auto =
+      AristoApiProfGetFilUbeFn.profileRunner:
+        result = api.getFilUbe(a, b)
+
   profApi.getKeyRc =
     proc(a: AristoDbRef; b: VertexID): auto =
       AristoApiProfGetKeyRcFn.profileRunner:

nimbus/db/aristo/aristo_check.nim

@@ -20,7 +20,7 @@ import
   results,
   ./aristo_walk/persistent,
   "."/[aristo_desc, aristo_get, aristo_init, aristo_utils],
-  ./aristo_check/[check_be, check_top]
+  ./aristo_check/[check_be, check_journal, check_top]
 
 # ------------------------------------------------------------------------------
 # Public functions
@@ -55,7 +55,7 @@ proc checkBE*(
     cache = true;                      # Also verify against top layer cache
     fifos = false;                     # Also verify cascaded filter fifos
       ): Result[void,(VertexID,AristoError)] =
-  ## Veryfy database backend structure. If the argument `relax` is set `false`,
+  ## Verify database backend structure. If the argument `relax` is set `false`,
   ## all necessary Merkle hashes are compiled and verified. If the argument
   ## `cache` is set `true`, the cache is also checked so that a safe operation
   ## (like `resolveBackendFilter()`) will leave the backend consistent.
@@ -79,6 +79,18 @@ proc checkBE*(
   of BackendVoid:
     return VoidBackendRef.checkBE(db, cache=cache, relax=relax)
 
+proc checkJournal*(
+    db: AristoDbRef;                   # Database, top layer
+      ): Result[void,(QueueID,AristoError)] =
+  ## Verify database backend journal.
+  case db.backend.kind:
+  of BackendMemory:
+    return MemBackendRef.checkJournal(db)
+  of BackendRocksDB:
+    return RdbBackendRef.checkJournal(db)
+  of BackendVoid:
+    return ok() # no journal
+
 
 proc check*(
     db: AristoDbRef;                   # Database, top layer
@@ -89,7 +101,11 @@ proc check*(
       ): Result[void,(VertexID,AristoError)] =
   ## Shortcut for running `checkTop()` followed by `checkBE()`
   ? db.checkTop(proofMode = proofMode)
-  ? db.checkBE(relax = relax, cache = cache)
+  ? db.checkBE(relax = relax, cache = cache, fifos = fifos)
+  if fifos:
+    let rc = db.checkJournal()
+    if rc.isErr:
+      return err((VertexID(0),rc.error[1]))
   ok()
 
 # ------------------------------------------------------------------------------

nimbus/db/aristo/aristo_check/check_be.nim

@@ -185,7 +185,7 @@ proc checkBE*[T: RdbBackendRef|MemBackendRef|VoidBackendRef](
     if fifos and
        not db.backend.isNil and
        not db.backend.journal.isNil:
-      var lastTrg = db.getKeyUBE(VertexID(1)).get(otherwise = VOID_HASH_KEY)
+      var lastTrg = db.getKeyUbe(VertexID(1)).get(otherwise = VOID_HASH_KEY)
                       .to(Hash256)
       for (qid,filter) in db.backend.T.walkFifoBe: # walk in fifo order
         if filter.src != lastTrg:

nimbus/db/aristo/aristo_check/check_journal.nim

@@ -0,0 +1,203 @@
+# nimbus-eth1
+# Copyright (c) 2023-2024 Status Research & Development GmbH
+# Licensed under either of
+#  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or
+#    http://www.apache.org/licenses/LICENSE-2.0)
+#  * MIT license ([LICENSE-MIT](LICENSE-MIT) or
+#    http://opensource.org/licenses/MIT)
+# at your option. This file may not be copied, modified, or distributed
+# except according to those terms.
+
+{.push raises: [].}
+
+import
+  std/[algorithm, sequtils, sets, tables],
+  eth/common,
+  results,
+  ../aristo_filter/filter_scheduler,
+  ../aristo_walk/persistent,
+  ".."/[aristo_desc, aristo_blobify]
+
+const
+  ExtraDebugMessages = false
+
+type
+  JrnRec = tuple
+    src: Hash256
+    trg: Hash256
+    size: int
+
+when ExtraDebugMessages:
+  import
+    ../aristo_debug
+
+# ------------------------------------------------------------------------------
+# Private functions and helpers
+# ------------------------------------------------------------------------------
+
+template noValueError(info: static[string]; code: untyped) =
+  try:
+    code
+  except ValueError as e:
+    raiseAssert info & ", name=\"" & $e.name & "\", msg=\"" & e.msg & "\""
+
+when ExtraDebugMessages:
+  proc pp(t: var Table[QueueID,JrnRec]): string =
+    result = "{"
+    for qid in t.keys.toSeq.sorted:
+      t.withValue(qid,w):
+        result &= qid.pp & "#" & $w[].size & ","
+    if result[^1] == '{':
+      result &= "}"
+    else:
+      result[^1] = '}'
+
+  proc pp(t: seq[QueueID]): string =
+    result = "{"
+    var list = t
+    for n in 2 ..< list.len:
+      if list[n-1] == list[n] - 1 and
+         (list[n-2] == QueueID(0) or list[n-2] == list[n] - 2):
+        list[n-1] = QueueID(0)
+    for w in list:
+      if w != QueueID(0):
+        result &= w.pp & ","
+      elif result[^1] == ',':
+        result[^1] = '.'
+        result &= "."
+    if result[^1] == '{':
+      result &= "}"
+    else:
+      result[^1] = '}'
+
+  proc pp(t: HashSet[QueueID]): string =
+    result = "{"
+    var list = t.toSeq.sorted
+    for n in 2 ..< list.len:
+      if list[n-1] == list[n] - 1 and
+         (list[n-2] == QueueID(0) or list[n-2] == list[n] - 2):
+        list[n-1] = QueueID(0)
+    for w in list:
+      if w != QueueID(0):
+        result &= w.pp & ","
+      elif result[^1] == ',':
+        result[^1] = '.'
+        result &= "."
+    if result[^1] == '{':
+      result &= "}"
+    else:
+      result[^1] = '}'
+
+# ------------------------------------------------------------------------------
+# Public functions
+# ------------------------------------------------------------------------------
+
+proc checkJournal*[T: RdbBackendRef|MemBackendRef](
+    _: type T;
+    db: AristoDbRef;
+      ): Result[void,(QueueID,AristoError)] =
+  let jrn = db.backend.journal
+  if jrn.isNil: return ok()
+
+  var
+    nToQid: seq[QueueID]             # qids sorted by history/age
+    cached: HashSet[QueueID]         # `nToQid[]` as set
+    saved: Table[QueueID,JrnRec]
+    error: (QueueID,AristoError)
+
+  when ExtraDebugMessages:
+    var
+      sizeTally = 0
+      maxBlock = 0
+
+    proc moan(n = -1, s = "", listOk = true) =
+      var txt = ""
+      if 0 <= n:
+        txt &= " (" & $n & ")"
+      if error[1] != AristoError(0):
+        txt &= " oops"
+      txt &=
+        " jLen=" & $jrn.len &
+        " tally=" & $sizeTally &
+        " maxBlock=" & $maxBlock &
+        ""
+      if 0 < s.len:
+        txt &= " " & s
+      if error[1] != AristoError(0):
+        txt &=
+          " errQid=" &  error[0].pp &
+          " error=" &  $error[1] &
+          ""
+      if listOk:
+        txt &=
+          "\n    cached=" &  cached.pp &
+          "\n    saved=" & saved.pp &
+          ""
+      debugEcho "*** checkJournal", txt
+  else:
+    template moan(n = -1, s = "", listOk = true) =
+      discard
+
+  # Collect cached handles
+  for n in 0 ..< jrn.len:
+    let qid = jrn[n]
+    # Must be no overlap
+    if qid in cached:
+      error = (qid,CheckJrnCachedQidOverlap)
+      moan(2)
+      return err(error)
+    cached.incl qid
+    nToQid.add qid
+
+  # Collect saved data
+  for (qid,fil) in db.backend.T.walkFilBe():
+    var jrnRec: JrnRec
+    jrnRec.src = fil.src
+    jrnRec.trg = fil.trg
+
+    when ExtraDebugMessages:
+      let rc = fil.blobify
+      if rc.isErr:
+        moan(5)
+        return err((qid,rc.error))
+      jrnRec.size = rc.value.len
+      if maxBlock < jrnRec.size:
+        maxBlock = jrnRec.size
+      sizeTally += jrnRec.size
+
+    saved[qid] = jrnRec
+
+  # Compare cached against saved data
+  let
+    savedQids = saved.keys.toSeq.toHashSet
+    unsavedQids = cached - savedQids
+    staleQids = savedQids - cached
+
+  if 0 < unsavedQids.len:
+    error = (unsavedQids.toSeq.sorted[0],CheckJrnSavedQidMissing)
+    moan(6)
+    return err(error)
+
+  if 0 < staleQids.len:
+    error = (staleQids.toSeq.sorted[0], CheckJrnSavedQidStale)
+    moan(7)
+    return err(error)
+
+  # Compare whether journal records link together
+  if 1 < nToQid.len:
+    noValueError("linked journal records"):
+      var prvRec = saved[nToQid[0]]
+      for n in 1 ..< nToQid.len:
+        let thisRec = saved[nToQid[n]]
+        if prvRec.trg != thisRec.src:
+          error = (nToQid[n],CheckJrnLinkingGap)
+          moan(8, "qidInx=" & $n)
+          return err(error)
+        prvRec = thisRec
+
+  moan(9, listOk=false)
+  ok()
+
+# ------------------------------------------------------------------------------
+# End
+# ------------------------------------------------------------------------------

nimbus/db/aristo/aristo_constants.nim

@@ -41,10 +41,21 @@ const
     ## Useful shortcut
 
   DEFAULT_QID_QUEUES* = [
-    (128,   0), ## Consecutive list of 128 filter slots
+    (128, 0), # Consecutive list of (at least) 128 filter slots
-    ( 64,  63), ## Overflow list, 64 filters, skipping 63 filters in-between
+    ( 16, 3), # Overflow list with (at least) 16 filter slots (with gap size 3)
-    ( 64, 127), ## ..
+              # each slot covering 4 filters from previous list
-    ( 64, 255)]
+    (  1, 1), # ..
+    (  1, 1)]
+    ## The `DEFAULT_QID_QUEUES` schedule has the following properties:
+    ## * most recent consecutive slots: 128
+    ## * maximal slots used: 151
+    ## * covered backlog savings: between 216..231
+    ## This was calculated via the `capacity()` function from the
+    ## `filter_scheduler.nim` source. So, saving each block after executing
+    ## it, the previous 128 block chain states will be directly accessible.
+    ## For older block chain states (of at least back to 216), the system can
+    ## be positioned before the desired state and block by block executed
+    ## forward.
 
   SUB_TREE_DISPOSAL_MAX* = 200_000
     ## Some limit for disposing sub-trees in one go using `delete()`.

nimbus/db/aristo/aristo_desc/desc_error.nim

@@ -169,6 +169,12 @@ type
     CheckBeFifoSrcTrgMismatch
     CheckBeFifoTrgNotStateRoot
 
+    # Jornal check `checkJournal()`
+    CheckJrnCachedQidOverlap
+    CheckJrnSavedQidMissing
+    CheckJrnSavedQidStale
+    CheckJrnLinkingGap
+
     # Neighbour vertex, tree traversal `nearbyRight()` and `nearbyLeft()`
     NearbyBeyondRange
     NearbyBranchError

nimbus/db/aristo/aristo_desc/desc_structural.nim

@@ -324,7 +324,7 @@ func to*(node: NodeRef; T: type VertexRef): T =
   node.VertexRef.dup
 
 func to*(a: array[4,tuple[size, width: int]]; T: type QidLayoutRef): T =
-  ## Convert a size-width array to a `QidLayoutRef` layout. Over large
+  ## Convert a size-width array to a `QidLayoutRef` layout. Overly large
   ## array field values are adjusted to its maximal size.
   var q: array[4,QidSpec]
   for n in 0..3:
@@ -335,18 +335,20 @@ func to*(a: array[4,tuple[size, width: int]]; T: type QidLayoutRef): T =
   T(q: q)
 
 func to*(a: array[4,tuple[size, width, wrap: int]]; T: type QidLayoutRef): T =
-  ## Convert a size-width-wrap array to a `QidLayoutRef` layout. Over large
+  ## Convert a size-width-wrap array to a `QidLayoutRef` layout. Overly large
   ## array field values are adjusted to its maximal size. Too small `wrap`
-  ## values are adjusted to its minimal size.
+  ## field values are adjusted to its minimal size.
   var q: array[4,QidSpec]
   for n in 0..2:
     q[n] = (min(a[n].size.uint, QidSpecSizeMax),
             min(a[n].width.uint, QidSpecWidthMax),
-            QueueID(max(a[n].size + a[n+1].width, a[n].width+1, a[n].wrap)))
+            QueueID(max(a[n].size + a[n+1].width, a[n].width+1,
+                        min(a[n].wrap, DefaultQidWrap.int))))
   q[0].width = 0
   q[3] = (min(a[3].size.uint, QidSpecSizeMax),
           min(a[3].width.uint, QidSpecWidthMax),
-          QueueID(max(a[3].size, a[3].width, a[3].wrap)))
+          QueueID(max(a[3].size, a[3].width,
+                      min(a[3].wrap, DefaultQidWrap.int))))
   T(q: q)
 
 # ------------------------------------------------------------------------------

nimbus/db/aristo/aristo_filter.nim

@@ -73,7 +73,7 @@ proc merge*(
   ## argument `filter`, all the `top` and `stack` layers should be cleared.
   ##
   let ubeRoot = block:
-    let rc = db.getKeyUBE VertexID(1)
+    let rc = db.getKeyUbe VertexID(1)
     if rc.isOk:
       rc.value.to(Hash256)
     elif rc.error == GetKeyNotFound:

nimbus/db/aristo/aristo_filter/filter_fifos.nim

@@ -75,6 +75,7 @@ proc fifosStore*(
   # Update journal filters and calculate database update
   var
     instr = FifoInstr(scd: upd.fifo)
+    dbClear: seq[QueueID]
     hold: seq[FilterRef]
     saved = false
 
@@ -100,10 +101,12 @@ proc fifosStore*(
 
     of HoldQid:
       # Push filter
+      dbClear.add act.qid
       hold.add be.getFilterOrReturn act.qid
 
       # Merge additional journal filters into top filter
       for w in act.qid+1 .. act.xid:
+        dbClear.add w
         let lower = be.getFilterOrReturn w
         hold[^1] = hold[^1].joinFiltersOrReturn lower
 
@@ -115,6 +118,9 @@ proc fifosStore*(
         let upper = hold.pop
         lower = upper.joinFiltersOrReturn lower
       instr.put.add (act.qid, lower)
+      for qid in dbClear:
+        instr.put.add (qid, FilterRef(nil))
+      dbClear.setLen(0)
 
   if not saved:
     return err(FilExecSaveMissing)

nimbus/db/aristo/aristo_filter/filter_helpers.nim

@@ -127,8 +127,8 @@ proc getFilterOverlap*(
   ## Return the number of journal filters in the leading chain that is
   ## reverted by the argument `filter`. A heuristc approach is used here
   ## for an argument `filter` with a valid filter ID when the chain is
-  ## longer than one items. So only single chain overlaps a guaranteed to
+  ## longer than one items. Only single step filter overlaps are guaranteed
-  ## be found.
+  ## to be found.
   ##
   # Check against the top-fifo entry
   let qid = be.journal[0]
@@ -148,7 +148,7 @@ proc getFilterOverlap*(
   if filter.trg == top.trg:
     return 1
 
-  # Check against sme stored filter IDs
+  # Check against some stored filter IDs
   if filter.isValid:
     let rc = be.getFilterFromFifo(filter.fid, earlierOK=true)
     if rc.isOk:

nimbus/db/aristo/aristo_filter/filter_merge.nim

@@ -75,7 +75,7 @@ proc merge*(
     if vtx.isValid or not newFilter.sTab.hasKey vid:
       newFilter.sTab[vid] = vtx
     elif newFilter.sTab.getOrVoid(vid).isValid:
-      let rc = db.getVtxUBE vid
+      let rc = db.getVtxUbe vid
       if rc.isOk:
         newFilter.sTab[vid] = vtx # VertexRef(nil)
       elif rc.error == GetVtxNotFound:
@@ -87,7 +87,7 @@ proc merge*(
     if key.isValid or not newFilter.kMap.hasKey vid:
       newFilter.kMap[vid] = key
     elif newFilter.kMap.getOrVoid(vid).isValid:
-      let rc = db.getKeyUBE vid
+      let rc = db.getKeyUbe vid
       if rc.isOk:
         newFilter.kMap[vid] = key
       elif rc.error == GetKeyNotFound:

nimbus/db/aristo/aristo_filter/filter_reverse.nim

@@ -1,5 +1,5 @@
 # nimbus-eth1
-# Copyright (c) 2023 Status Research & Development GmbH
+# Copyright (c) 2023-2024 Status Research & Development GmbH
 # Licensed under either of
 #  * Apache License, version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or
 #    http://www.apache.org/licenses/LICENSE-2.0)
@@ -34,7 +34,7 @@ proc revFilter*(
 
   # Get vid generator state on backend
   block:
-    let rc = db.getIdgUBE()
+    let rc = db.getIdgUbe()
     if rc.isOk:
       rev.vGen = rc.value
     elif rc.error != GetIdgNotFound:
@@ -42,7 +42,7 @@ proc revFilter*(
 
   # Calculate reverse changes for the `sTab[]` structural table
   for vid in filter.sTab.keys:
-    let rc = db.getVtxUBE vid
+    let rc = db.getVtxUbe vid
     if rc.isOk:
       rev.sTab[vid] = rc.value
     elif rc.error == GetVtxNotFound:
@@ -52,7 +52,7 @@ proc revFilter*(
 
   # Calculate reverse changes for the `kMap` sequence.
   for vid in filter.kMap.keys:
-    let rc = db.getKeyUBE vid
+    let rc = db.getKeyUbe vid
     if rc.isOk:
       rev.kMap[vid] = rc.value
     elif rc.error == GetKeyNotFound:

nimbus/db/aristo/aristo_filter/filter_scheduler.nim

@@ -261,7 +261,7 @@ func capacity(
 func capacity*(
     ctx: openArray[tuple[size, width, wrap: int]]; # Schedule layout
       ): tuple[maxQueue: int, minCovered: int, maxCovered: int] =
-  ## Variant of `capacity()` below.
+  ## Variant of `capacity()`.
   ctx.toSeq.mapIt((it[0],it[1])).capacity
 
 func capacity*(
@@ -289,16 +289,17 @@ func addItem*(
   ##    SaveQid <queue-id>         -- Store a new item under the address
   ##                               -- <queue-id> on the database.
   ##
-  ##    HoldQid <from-id>..<to-id> -- Move the records accessed by the argument
+  ##    HoldQid <from-id>..<to-id> -- Move the records referred to by the
-  ##                               -- addresses from the database to the right
+  ##                               -- argument addresses from the database to
-  ##                               -- end of the local hold queue. The age of
+  ##                               -- the right end of the local hold queue.
-  ##                               -- the items on the hold queue increases
+  ##                               -- The age of the items on the hold queue
-  ##                               -- left to right.
+  ##                               -- increases left to right.
   ##
   ##    DequQid <queue-id>         -- Merge items from the hold queue into a
   ##                               -- new item and store it under the address
   ##                               -- <queue-id> on the database. Clear the
-  ##                               -- the hold queue.
+  ##                               -- the hold queue and the corresponding
+  ##                               -- items on the database.
   ##
   ##    DelQid <queue-id>          -- Delete item. This happens if the last
   ##                               -- oberflow queue needs to make space for
@@ -557,7 +558,7 @@ func `[]`*(
     bix: BackwardsIndex;                           # Index into latest items
       ): QueueID =
   ## Variant of `[]` for provifing `[^bix]`.
-  fifo[fifo.state.len - bix.distinctBase]
+  fifo[fifo.len - bix.distinctBase]
 
 
 func `[]`*(

nimbus/db/aristo/aristo_get.nim

@@ -22,7 +22,7 @@ import
 # Public functions
 # ------------------------------------------------------------------------------
 
-proc getIdgUBE*(
+proc getIdgUbe*(
     db: AristoDbRef;
       ): Result[seq[VertexID],AristoError] =
   ## Get the ID generator state from the unfiltered backened if available.
@@ -31,7 +31,7 @@ proc getIdgUBE*(
     return be.getIdgFn()
   err(GetIdgNotFound)
 
-proc getFqsUBE*(
+proc getFqsUbe*(
     db: AristoDbRef;
       ): Result[seq[(QueueID,QueueID)],AristoError] =
   ## Get the list of filter IDs unfiltered backened if available.
@@ -40,7 +40,7 @@ proc getFqsUBE*(
     return be.getFqsFn()
   err(GetFqsNotFound)
 
-proc getVtxUBE*(
+proc getVtxUbe*(
     db: AristoDbRef;
     vid: VertexID;
       ): Result[VertexRef,AristoError] =
@@ -50,17 +50,17 @@ proc getVtxUBE*(
     return be.getVtxFn vid
   err GetVtxNotFound
 
-proc getKeyUBE*(
+proc getKeyUbe*(
     db: AristoDbRef;
     vid: VertexID;
       ): Result[HashKey,AristoError] =
-  ## Get the merkle hash/key from the unfiltered backend if available.
+  ## Get the Merkle hash/key from the unfiltered backend if available.
   let be = db.backend
   if not be.isNil:
     return be.getKeyFn vid
   err GetKeyNotFound
 
-proc getFilUBE*(
+proc getFilUbe*(
     db: AristoDbRef;
     qid: QueueID;
       ): Result[FilterRef,AristoError] =
@@ -78,7 +78,7 @@ proc getIdgBE*(
   ## Get the ID generator state the `backened` layer if available.
   if not db.roFilter.isNil:
     return ok(db.roFilter.vGen)
-  db.getIdgUBE()
+  db.getIdgUbe()
 
 proc getVtxBE*(
     db: AristoDbRef;
@@ -90,7 +90,7 @@ proc getVtxBE*(
     if vtx.isValid:
       return ok(vtx)
     return err(GetVtxNotFound)
-  db.getVtxUBE vid
+  db.getVtxUbe vid
 
 proc getKeyBE*(
     db: AristoDbRef;
@@ -102,7 +102,7 @@ proc getKeyBE*(
     if key.isValid:
       return ok(key)
     return err(GetKeyNotFound)
-  db.getKeyUBE vid
+  db.getKeyUbe vid
 
 # ------------------
 

nimbus/db/aristo/aristo_tx.nim

@@ -279,7 +279,7 @@ proc forkWith*(
 
   # Try `(vid,key)` on unfiltered backend
   block:
-    let beKey = db.getKeyUBE(vid).valueOr: VOID_HASH_KEY
+    let beKey = db.getKeyUbe(vid).valueOr: VOID_HASH_KEY
     if beKey == key:
       let rc = db.fork(noFilter = true)
       if rc.isOk:
@@ -451,7 +451,7 @@ proc stow*(
     if db.roFilter.isValid:
       db.top.final.vGen = db.roFilter.vGen
     else:
-      let rc = db.getIdgUBE()
+      let rc = db.getIdgUbe()
       if rc.isOk:
         db.top.final.vGen = rc.value
       else:

nimbus/db/core_db/backend/aristo_db.nim

@@ -242,6 +242,11 @@ func toAristo*(mBe: CoreDbMptBackendRef): AristoDbRef =
   if not mBe.isNil and mBe.parent.isAristo:
     return mBe.AristoCoreDbMptBE.adb
 
+proc toAristoOldestStateRoot*(mBe: CoreDbMptBackendRef): Hash256 =
+  if not mBe.isNil and mBe.parent.isAristo:
+    return mBe.parent.AristoCoreDbRef.adbBase.toJournalOldestStateRoot()
+  EMPTY_ROOT_HASH
+
 # ------------------------------------------------------------------------------
 # Public aristo iterators
 # ------------------------------------------------------------------------------

nimbus/db/core_db/backend/aristo_db/handlers_aristo.nim

@@ -17,6 +17,7 @@ import
   stew/byteutils,
   results,
   ../../../aristo,
+  ../../../aristo/aristo_filter/filter_scheduler,
   ../../base,
   ../../base/base_desc,
   ./common_desc
@@ -566,6 +567,20 @@ func toVoidRc*[T](
     return ok()
   err((VoidVID,rc.error).toError(base, info, error))
 
+proc toJournalOldestStateRoot*(base: AristoBaseRef): Hash256 =
+  let
+    adb = base.ctx.mpt
+    be = adb.backend
+  if not be.isNil:
+    let jrn = be.journal
+    if not jrn.isNil:
+      let qid = jrn[^1]
+      if qid.isValid:
+        let rc = base.api.getFilUbe(adb, qid)
+        if rc.isOk:
+          return rc.value.trg
+  EMPTY_ROOT_HASH
+
 # ---------------------
 
 func to*(dsc: CoreDxMptRef, T: type AristoDbRef): T =

nimbus/db/core_db/core_apps_newapi.nim

@@ -18,9 +18,10 @@ import
   chronicles,
   eth/[common, rlp],
   results,
-  stew/byteutils,
+  stew/[byteutils, endians2],
   "../.."/[errors, constants],
   ".."/[aristo, storage_types],
+  ./backend/aristo_db,
   "."/base
 
 logScope:
@@ -88,6 +89,23 @@ template discardRlpException(info: static[string]; code: untyped) =
   except RlpError as e:
     warn logTxt info, error=($e.name), msg=e.msg
 
+# ---------
+
+func to(bn: BlockNumber; T: type Blob): T =
+  if bn <= high(uint64).toBlockNumber:
+    bn.truncate(uint64).toBytesBE.toSeq
+  else:
+    bn.toBytesBE.toSeq
+
+func to(data: openArray[byte]; T: type BlockNumber): T =
+  case data.len:
+  of 8:
+    return uint64.fromBytesBE(data).toBlockNumber
+  of 32:
+    return UInt256.fromBytesBE(data).toBlockNumber
+  else:
+    discard
+
 # ------------------------------------------------------------------------------
 # Private iterators
 # ------------------------------------------------------------------------------
@@ -320,7 +338,6 @@ proc markCanonicalChain(
 
   return true
 
-
 # ------------------------------------------------------------------------------
 # Public functions
 # ------------------------------------------------------------------------------
@@ -330,6 +347,27 @@ proc exists*(db: CoreDbRef, hash: Hash256): bool =
     warn logTxt "exisis()", hash, action="hasKey()", error=($$error)
     return false
 
+proc getBlockNumber*(db: CoreDbRef; stateRoot: Hash256): BlockNumber =
+  const info = "getBlockNumber()"
+  if stateRoot != EMPTY_ROOT_HASH:
+    let
+      kvt = db.newKvt()
+      data = kvt.get(stRootToBlockNumKey(stateRoot).toOpenArray).valueOr:
+        if error.error != KvtNotFound:
+          warn logTxt info, stateRoot, action="get()", error=($$error)
+        return
+    return data.to(BlockNumber)
+
+proc getOldestJournalBlockNumber*(db: CoreDbRef): BlockNumber =
+  ## Returns the block number implied by the database journal if there is any,
+  ## or `BlockNumber(0)`. At the moment, only the `Aristo` database has a
+  ## journal.
+  ##
+  let be = db.ctx.getMpt(CtGeneric).backend
+  if be.parent.isAristo:
+    return db.getBlockNumber be.toAristoOldestStateRoot()
+
+
 proc getBlockHeader*(
     db: CoreDbRef;
     blockHash: Hash256;
@@ -523,8 +561,18 @@ proc getAncestorsHashes*(
     dec ancestorCount
 
 proc addBlockNumberToHashLookup*(db: CoreDbRef; header: BlockHeader) =
-  let blockNumberKey = blockNumberToHashKey(header.blockNumber)
+  ## The function stores lookup for
-  db.newKvt.put(blockNumberKey.toOpenArray, rlp.encode(header.hash)).isOkOr:
+  ## ::
+  ##   header.stateRoot -> header.blockNumber -> header.hash()
+  ##
+  let
+    blockNumberKey = blockNumberToHashKey(header.blockNumber)
+    stRootKey = stRootToBlockNumKey(header.stateRoot)
+    kvt = db.newKvt()
+  kvt.put(stRootKey.toOpenArray, header.blockNumber.to(Blob)).isOkOr:
+    warn logTxt "addBlockNumberToHashLookup()",
+      stRootKey, action="put()", `error`=($$error)
+  kvt.put(blockNumberKey.toOpenArray, rlp.encode(header.hash)).isOkOr:
     warn logTxt "addBlockNumberToHashLookup()",
       blockNumberKey, action="put()", error=($$error)
 
@@ -558,6 +606,26 @@ proc persistTransactions*(
       warn logTxt info, action="state()"
     return EMPTY_ROOT_HASH
 
+proc forgetHistory*(
+    db: CoreDbRef;
+    blockNum: BlockNumber;
+      ): bool
+      {.gcsafe, raises: [RlpError].} =
+  ## Remove all data related to the block number argument `num`. This function
+  ## returns `true`, if some history was available and deleted.
+  var blockHash: Hash256
+  if db.getBlockHash(blockNum, blockHash):
+    let kvt = db.newKvt()
+    # delete blockNum->blockHash
+    discard kvt.del(blockNumberToHashKey(blockNum).toOpenArray)
+    result = true
+
+    var header: BlockHeader
+    if db.getBlockHeader(blockHash, header):
+      # delete blockHash->header, stateRoot->blockNum
+      discard kvt.del(genericHashKey(blockHash).toOpenArray)
+      discard kvt.del(stRootToBlockNumKey(header.stateRoot).toOpenArray)
+
 proc getTransaction*(
     db: CoreDbRef;
     txRoot: Hash256;
@@ -858,6 +926,12 @@ proc persistHeaderToDbWithoutSetHead*(
   let
     kvt = db.newKvt()
     scoreKey = blockHashToScoreKey(headerHash)
+
+  # This extra call `addBlockNumberToHashLookup()` has been added in order
+  # to access the burrent block by the state root. So it can be deleted
+  # if not needed, anymore.
+  db.addBlockNumberToHashLookup(header)
+
   kvt.put(scoreKey.toOpenArray, rlp.encode(score)).isOkOr:
     warn logTxt "persistHeaderToDbWithoutSetHead()",
       scoreKey, action="put()", `error`=($$error)

nimbus/db/core_db/memory_only.nim

@@ -34,6 +34,7 @@ export
   isAristo,
   toAristo,
   toAristoProfData,
+  toAristoOldestStateRoot,
 
   # see `legacy_db`
   isLegacy,

nimbus/db/storage_types.nim

@@ -32,6 +32,7 @@ type
     snapSyncStorageSlot
     snapSyncStateRoot
     blockHashToBlockWitness
+    stRootToBlockNum
 
   DbKey* = object
     # The first byte stores the key type. The rest are key-specific values
@@ -135,6 +136,11 @@ proc blockHashToBlockWitnessKey*(h: Hash256): DbKey {.inline.} =
   result.data[1 .. 32] = h.data
   result.dataEndPos = uint8 32
 
+proc stRootToBlockNumKey*(h: Hash256): DbKey =
+  result.data[0] = byte ord(stRootToBlockNum)
+  result.data[1 .. 32] = h.data
+  result.dataEndPos = uint8 32
+
 template toOpenArray*(k: DbKey): openArray[byte] =
   k.data.toOpenArray(0, int(k.dataEndPos))
 

tests/test_aristo.nim

@@ -74,8 +74,9 @@ proc setErrorLevel {.used.} =
 
 proc miscRunner(
     noisy = true;
-    qidSampleSize = QidSample;
+    layout = LyoSamples[0];
-     ) =
+      ) =
+  let (lyo,qidSampleSize) = layout
 
   suite "Aristo: Miscellaneous tests":
 
@@ -83,10 +84,10 @@ proc miscRunner(
       check noisy.testVidRecycleLists()
 
     test &"Low level cascaded fifos API (sample size: {qidSampleSize})":
-      check noisy.testQidScheduler(sampleSize = qidSampleSize)
+      check noisy.testQidScheduler(layout = lyo, sampleSize = qidSampleSize)
 
     test &"High level cascaded fifos API (sample size: {qidSampleSize})":
-      check noisy.testFilterFifo(sampleSize = qidSampleSize)
+      check noisy.testFilterFifo(layout = lyo, sampleSize = qidSampleSize)
 
     test "Short keys and other patholgical cases":
       check noisy.testShortKeys()
@@ -197,7 +198,8 @@ when isMainModule:
     noisy.accountsRunner(persistent=false)
 
   when true: # and false:
-    noisy.miscRunner(qidSampleSize = 1_000)
+    for n,w in LyoSamples:
+      noisy.miscRunner() # layouts = (w[0], 1_000))
 
   # This one uses dumps from the external `nimbus-eth1-blob` repo
   when true and false:

tests/test_aristo/test_filter.nim

@@ -64,7 +64,10 @@ proc fifos(be: BackendRef): seq[seq[(QueueID,FilterRef)]] =
     discard
   check be.kind == BackendMemory or be.kind == BackendRocksDB
 
-func flatten(a: seq[seq[(QueueID,FilterRef)]]): seq[(QueueID,FilterRef)] {.used.} =
+func flatten(
+    a: seq[seq[(QueueID,FilterRef)]];
+      ): seq[(QueueID,FilterRef)]
+      {.used.} =
   for w in a:
     result &= w
 
@@ -238,7 +241,7 @@ proc isDbEq(a, b: FilterRef; db: AristoDbRef; noisy = true): bool =
         if aVtx.isValid and bVtx.isValid:
           return false
         # The valid one must match the backend data
-        let rc = db.getVtxUBE vid
+        let rc = db.getVtxUbe vid
         if rc.isErr:
           return false
         let vtx = if aVtx.isValid: aVtx else: bVtx
@@ -246,7 +249,7 @@ proc isDbEq(a, b: FilterRef; db: AristoDbRef; noisy = true): bool =
           return false
 
       elif not vid.isValid and not bTab.hasKey vid:
-        let rc = db.getVtxUBE vid
+        let rc = db.getVtxUbe vid
         if rc.isOk:
           return false # Exists on backend but missing on `bTab[]`
         elif rc.error != GetKeyNotFound:
@@ -268,7 +271,7 @@ proc isDbEq(a, b: FilterRef; db: AristoDbRef; noisy = true): bool =
         if aKey.isValid and bKey.isValid:
           return false
         # The valid one must match the backend data
-        let rc = db.getKeyUBE vid
+        let rc = db.getKeyUbe vid
         if rc.isErr:
           return false
         let key = if aKey.isValid: aKey else: bKey
@@ -276,7 +279,7 @@ proc isDbEq(a, b: FilterRef; db: AristoDbRef; noisy = true): bool =
           return false
 
       elif not vid.isValid and not bMap.hasKey vid:
-        let rc = db.getKeyUBE vid
+        let rc = db.getKeyUbe vid
         if rc.isOk:
           return false # Exists on backend but missing on `bMap[]`
         elif rc.error != GetKeyNotFound:
@@ -348,18 +351,24 @@ proc checkBeOk(
     dx: DbTriplet;
     relax = false;
     forceCache = false;
+    fifos = true;
     noisy = true;
       ): bool =
   ## ..
   for n in 0 ..< dx.len:
-    let
+    let cache = if forceCache: true else: dx[n].dirty.len == 0
-      cache = if forceCache: true else: dx[n].dirty.len == 0
+    block:
-      rc = dx[n].checkBE(relax=relax, cache=cache)
+      let rc = dx[n].checkBE(relax=relax, cache=cache, fifos=fifos)
-    xCheckRc rc.error == (0,0):
+      xCheckRc rc.error == (0,0):
-      noisy.say "***", "db check failed",
+        noisy.say "***", "db checkBE failed",
-        " n=", n, "/", dx.len-1,
+          " n=", n, "/", dx.len-1,
-        " cache=", cache
+          " cache=", cache
-
+    if fifos:
+      let rc = dx[n].checkJournal()
+      xCheckRc rc.error == (0,0):
+        noisy.say "***", "db checkJournal failed",
+          " n=", n, "/", dx.len-1,
+          " cache=", cache
   true
 
 proc checkFilterTrancoderOk(
@@ -602,7 +611,7 @@ proc testDistributedAccess*(
 
       # Check/verify backends
       block:
-        let ok = dx.checkBeOk(noisy=noisy)
+        let ok = dx.checkBeOk(noisy=noisy,fifos=true)
         xCheck ok:
           noisy.say "*** testDistributedAccess (4)", "n=", n, "db3".dump db3
       block:
@@ -661,7 +670,7 @@ proc testDistributedAccess*(
 
       # Check/verify backends
       block:
-        let ok = dy.checkBeOk(noisy=noisy)
+        let ok = dy.checkBeOk(noisy=noisy,fifos=true)
         xCheck ok
       block:
         let ok = dy.checkFilterTrancoderOk(noisy=noisy)
@@ -675,8 +684,8 @@ proc testDistributedAccess*(
 
 proc testFilterFifo*(
     noisy = true;
-    layout = QidSlotLyo;                   # Backend fifos layout
+    layout = LyoSamples[0][0];             # Backend fifos layout
-    sampleSize = QidSample;                # Synthetic filters generation
+    sampleSize = LyoSamples[0][1];         # Synthetic filters generation
     reorgPercent = 40;                     # To be deleted and re-filled
     rdbPath = "";                          # Optional Rocks DB storage directory
       ): bool =
@@ -710,11 +719,23 @@ proc testFilterFifo*(
 
   # -------------------
 
+  block:
+    let rc = db.checkJournal()
+    xCheckRc rc.error == (0,0)
+
   for n in 1 .. sampleSize:
-    let storeFilterOK = be.storeFilter(serial=n)
+    #let trigger = n in {7,8}
-    xCheck storeFilterOK
+    #if trigger: show(n, be.journal.addItem.exec)
-    let validateFifoOk = be.validateFifo(serial=n)
+    block:
-    xCheck validateFifoOk
+      let storeFilterOK = be.storeFilter(serial=n)
+      xCheck storeFilterOK
+    block:
+      #if trigger: show(n)
+      let rc = db.checkJournal()
+      xCheckRc rc.error == (0,0)
+    block:
+      let validateFifoOk = be.validateFifo(serial=n)
+      xCheck validateFifoOk
 
   # Squash some entries on the fifo
   block:
@@ -739,6 +760,9 @@ proc testFilterFifo*(
     #show(n)
     let validateFifoOk = be.validateFifo(serial=n)
     xCheck validateFifoOk
+    block:
+      let rc = db.checkJournal()
+      xCheckRc rc.error == (0,0)
 
   true
 
@@ -746,7 +770,7 @@ proc testFilterFifo*(
 proc testFilterBacklog*(
     noisy: bool;
     list: openArray[ProofTrieData];        # Sample data for generating filters
-    layout = QidSlotLyo;                   # Backend fifos layout
+    layout = LyoSamples[0][0];             # Backend fifos layout
     reorgPercent = 40;                     # To be deleted and re-filled
     rdbPath = "";                          # Optional Rocks DB storage directory
     sampleSize = 777;                      # Truncate `list`
@@ -786,6 +810,9 @@ proc testFilterBacklog*(
     block:
       let rc = db.stow(persistent=true)
       xCheckRc rc.error == 0
+    block:
+      let rc = db.checkJournal()
+      xCheckRc rc.error == (0,0)
     let validateFifoOk = be.validateFifo(serial=n, hashesOk=true)
     xCheck validateFifoOk
     when false: # or true:
@@ -845,6 +872,9 @@ proc testFilterBacklog*(
     block:
       let rc = xb.check(relax=false)
       xCheckRc rc.error == (0,0)
+    block:
+      let rc = db.checkJournal()
+      xCheckRc rc.error == (0,0)
 
     #show(episode, "testFilterBacklog (3)")
 

tests/test_aristo/test_helpers.nim

@@ -31,8 +31,12 @@ type
     kvpLst*: seq[LeafTiePayload]
 
 const
-  QidSlotLyo* = [(4,0,10),(3,3,10),(3,4,10),(3,5,10)]
+  samples = [
-  QidSample* = (3 * QidSlotLyo.capacity.minCovered) div 2
+    [      (4,0,10),      (3,3,10),      (3,4,10),      (3,5,10)],
+    [(2,0,high int),(1,1,high int),(1,1,high int),(1,1,high int)],
+  ]
+
+  LyoSamples* = samples.mapIt((it, (3 * it.capacity.minCovered) div 2))
 
 # ------------------------------------------------------------------------------
 # Private helpers
@@ -106,7 +110,7 @@ proc say*(noisy = false; pfx = "***"; args: varargs[string, `$`]) =
 func `==`*[T: AristoError|VertexID](a: T, b: int): bool =
   a == T(b)
 
-func `==`*(a: (VertexID,AristoError), b: (int,int)): bool =
+func `==`*(a: (VertexID|QueueID,AristoError), b: (int,int)): bool =
   (a[0].int,a[1].int) == b
 
 func `==`*(a: (VertexID,AristoError), b: (int,AristoError)): bool =

tests/test_aristo/test_misc.nim

@@ -335,8 +335,8 @@ proc testVidRecycleLists*(noisy = true; seed = 42): bool =
 
 proc testQidScheduler*(
     noisy = true;
-    layout = QidSlotLyo;
+    layout = LyoSamples[0][0];
-    sampleSize = QidSample;
+    sampleSize = LyoSamples[0][1];
     reorgPercent = 40
       ): bool =
   ##

tests/test_aristo/test_tx.nim

@@ -43,7 +43,7 @@ const
     ## Policy settig for `pack()`
 
 let
-  TxQidLyo = QidSlotLyo.to(QidLayoutRef)
+  TxQidLyo = LyoSamples[0][0].to(QidLayoutRef)
     ## Cascaded filter slots layout for testing
 
 # ------------------------------------------------------------------------------