2022-07-01 11:42:17 +00:00
|
|
|
# Nimbus - Types, data structures and shared utilities used in network sync
|
|
|
|
#
|
|
|
|
# Copyright (c) 2018-2021 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.
|
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
## Snap sync components tester and TDD environment
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
import
|
2022-08-24 13:44:18 +00:00
|
|
|
std/[algorithm, distros, hashes, math, os, sets,
|
2022-08-12 15:42:07 +00:00
|
|
|
sequtils, strformat, strutils, tables, times],
|
2022-07-01 11:42:17 +00:00
|
|
|
chronicles,
|
2022-11-25 14:56:42 +00:00
|
|
|
eth/[common, p2p, rlp],
|
|
|
|
eth/trie/[db, nibbles],
|
2022-08-12 15:42:07 +00:00
|
|
|
rocksdb,
|
2022-07-01 11:42:17 +00:00
|
|
|
stint,
|
2022-11-25 14:56:42 +00:00
|
|
|
stew/[byteutils, interval_set, results],
|
2022-07-01 11:42:17 +00:00
|
|
|
unittest2,
|
2022-08-12 15:42:07 +00:00
|
|
|
../nimbus/[chain_config, config, genesis],
|
|
|
|
../nimbus/db/[db_chain, select_backend, storage_types],
|
|
|
|
../nimbus/p2p/chain,
|
2022-08-24 13:44:18 +00:00
|
|
|
../nimbus/sync/types,
|
2022-08-04 08:04:30 +00:00
|
|
|
../nimbus/sync/snap/range_desc,
|
2022-10-14 16:40:32 +00:00
|
|
|
../nimbus/sync/snap/worker/db/[
|
2022-11-25 14:56:42 +00:00
|
|
|
hexary_desc, hexary_error, hexary_inspect, hexary_paths, rocky_bulk_load,
|
2022-11-16 23:51:06 +00:00
|
|
|
snapdb_accounts, snapdb_desc, snapdb_pivot, snapdb_storage_slots],
|
2022-08-12 15:42:07 +00:00
|
|
|
../nimbus/utils/prettify,
|
2022-09-02 18:16:09 +00:00
|
|
|
./replay/[pp, undump_blocks, undump_accounts, undump_storages],
|
|
|
|
./test_sync_snap/[bulk_test_xx, snap_test_xx, test_types]
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
const
|
|
|
|
baseDir = [".", "..", ".."/"..", $DirSep]
|
2022-08-12 15:42:07 +00:00
|
|
|
repoDir = [".", "tests"/"replay", "tests"/"test_sync_snap",
|
|
|
|
"nimbus-eth1-blobs"/"replay"]
|
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
# Reference file for finding the database directory
|
|
|
|
sampleDirRefFile = "sample0.txt.gz"
|
|
|
|
|
|
|
|
# Standard test samples
|
|
|
|
bChainCapture = bulkTest0
|
|
|
|
accSample = snapTest0
|
|
|
|
storSample = snapTest4
|
|
|
|
|
|
|
|
# Number of database slots (needed for timing tests)
|
2022-08-12 15:42:07 +00:00
|
|
|
nTestDbInstances = 9
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
type
|
|
|
|
TestDbs = object
|
2022-08-12 15:42:07 +00:00
|
|
|
## Provide enough spare empty databases
|
2022-07-01 11:42:17 +00:00
|
|
|
persistent: bool
|
|
|
|
dbDir: string
|
2022-08-12 15:42:07 +00:00
|
|
|
cdb: array[nTestDbInstances,ChainDb]
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
when defined(linux):
|
|
|
|
# The `detectOs(Ubuntu)` directive is not Windows compatible, causes an
|
|
|
|
# error when running the system command `lsb_release -d` in the background.
|
|
|
|
let isUbuntu32bit = detectOs(Ubuntu) and int.sizeof == 4
|
|
|
|
else:
|
|
|
|
const isUbuntu32bit = false
|
|
|
|
|
|
|
|
let
|
|
|
|
# Forces `check()` to print the error (as opposed when using `isOk()`)
|
2022-08-15 15:51:50 +00:00
|
|
|
OkHexDb = Result[void,HexaryDbError].ok()
|
2022-09-02 18:16:09 +00:00
|
|
|
OkStoDb = Result[void,seq[(int,HexaryDbError)]].ok()
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
# There was a problem with the Github/CI which results in spurious crashes
|
|
|
|
# when leaving the `runner()` if the persistent BaseChainDB initialisation
|
|
|
|
# was present, see `test_custom_network` for more details.
|
|
|
|
disablePersistentDB = isUbuntu32bit
|
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
var
|
|
|
|
xTmpDir: string
|
|
|
|
xDbs: TestDbs # for repeated storage/overwrite tests
|
|
|
|
xTab32: Table[ByteArray32,Blob] # extracted data
|
|
|
|
xTab33: Table[ByteArray33,Blob]
|
|
|
|
xVal32Sum, xVal32SqSum: float # statistics
|
|
|
|
xVal33Sum, xVal33SqSum: float
|
|
|
|
|
2022-07-01 11:42:17 +00:00
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
# Helpers
|
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
proc isOk(rc: ValidationResult): bool =
|
|
|
|
rc == ValidationResult.OK
|
|
|
|
|
2022-11-16 23:51:06 +00:00
|
|
|
proc isImportOk(rc: Result[SnapAccountsGaps,HexaryDbError]): bool =
|
|
|
|
if rc.isErr:
|
|
|
|
check rc.error == NothingSerious # prints an error if different
|
|
|
|
elif 0 < rc.value.innerGaps.len:
|
|
|
|
check rc.value.innerGaps == seq[NodeSpecs].default
|
|
|
|
else:
|
|
|
|
return true
|
|
|
|
|
2022-10-14 16:40:32 +00:00
|
|
|
proc toStoDbRc(r: seq[HexaryNodeReport]): Result[void,seq[(int,HexaryDbError)]]=
|
|
|
|
## Kludge: map error report to (older version) return code
|
|
|
|
if r.len != 0:
|
|
|
|
return err(r.mapIt((it.slot.get(otherwise = -1),it.error)))
|
|
|
|
ok()
|
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
proc findFilePath(file: string;
|
|
|
|
baseDir, repoDir: openArray[string]): Result[string,void] =
|
2022-07-01 11:42:17 +00:00
|
|
|
for dir in baseDir:
|
|
|
|
for repo in repoDir:
|
|
|
|
let path = dir / repo / file
|
|
|
|
if path.fileExists:
|
|
|
|
return ok(path)
|
2022-09-02 18:16:09 +00:00
|
|
|
echo "*** File not found \"", file, "\"."
|
2022-07-01 11:42:17 +00:00
|
|
|
err()
|
|
|
|
|
2022-08-24 13:44:18 +00:00
|
|
|
proc getTmpDir(sampleDir = sampleDirRefFile): string =
|
2022-08-12 15:42:07 +00:00
|
|
|
sampleDir.findFilePath(baseDir,repoDir).value.splitFile.dir
|
|
|
|
|
|
|
|
proc pp(d: Duration): string =
|
|
|
|
if 40 < d.inSeconds:
|
|
|
|
d.ppMins
|
|
|
|
elif 200 < d.inMilliseconds:
|
|
|
|
d.ppSecs
|
|
|
|
elif 200 < d.inMicroseconds:
|
|
|
|
d.ppMs
|
|
|
|
else:
|
|
|
|
d.ppUs
|
|
|
|
|
2022-08-15 15:51:50 +00:00
|
|
|
proc pp(rc: Result[Account,HexaryDbError]): string =
|
2022-08-12 15:42:07 +00:00
|
|
|
if rc.isErr: $rc.error else: rc.value.pp
|
2022-08-04 08:04:30 +00:00
|
|
|
|
2022-08-24 13:44:18 +00:00
|
|
|
proc pp(rc: Result[Hash256,HexaryDbError]): string =
|
|
|
|
if rc.isErr: $rc.error else: $rc.value.to(NodeTag)
|
|
|
|
|
2022-10-14 16:40:32 +00:00
|
|
|
proc pp(rc: Result[TrieNodeStat,HexaryDbError]; db: SnapDbBaseRef): string =
|
|
|
|
if rc.isErr: $rc.error else: rc.value.pp(db.hexaDb)
|
2022-09-16 07:24:12 +00:00
|
|
|
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
proc pp(a: NodeKey; collapse = true): string =
|
|
|
|
a.to(Hash256).pp(collapse)
|
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
proc ppKvPc(w: openArray[(string,int)]): string =
|
|
|
|
w.mapIt(&"{it[0]}={it[1]}%").join(", ")
|
2022-08-04 08:04:30 +00:00
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
proc say*(noisy = false; pfx = "***"; args: varargs[string, `$`]) =
|
|
|
|
if noisy:
|
|
|
|
if args.len == 0:
|
|
|
|
echo "*** ", pfx
|
|
|
|
elif 0 < pfx.len and pfx[^1] != ' ':
|
|
|
|
echo pfx, " ", args.toSeq.join
|
|
|
|
else:
|
|
|
|
echo pfx, args.toSeq.join
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
proc setTraceLevel =
|
|
|
|
discard
|
|
|
|
when defined(chronicles_runtime_filtering) and loggingEnabled:
|
|
|
|
setLogLevel(LogLevel.TRACE)
|
|
|
|
|
|
|
|
proc setErrorLevel =
|
|
|
|
discard
|
|
|
|
when defined(chronicles_runtime_filtering) and loggingEnabled:
|
|
|
|
setLogLevel(LogLevel.ERROR)
|
|
|
|
|
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
# Private functions
|
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
proc to(sample: AccountsSample; T: type seq[UndumpAccounts]): T =
|
|
|
|
## Convert test data into usable in-memory format
|
2022-08-24 13:44:18 +00:00
|
|
|
let file = sample.file.findFilePath(baseDir,repoDir).value
|
2022-09-02 18:16:09 +00:00
|
|
|
var root: Hash256
|
|
|
|
for w in file.undumpNextAccount:
|
|
|
|
let n = w.seenAccounts - 1
|
|
|
|
if n < sample.firstItem:
|
|
|
|
continue
|
|
|
|
if sample.lastItem < n:
|
|
|
|
break
|
|
|
|
if sample.firstItem == n:
|
|
|
|
root = w.root
|
|
|
|
elif w.root != root:
|
|
|
|
break
|
|
|
|
result.add w
|
|
|
|
|
|
|
|
proc to(sample: AccountsSample; T: type seq[UndumpStorages]): T =
|
|
|
|
## Convert test data into usable in-memory format
|
|
|
|
let file = sample.file.findFilePath(baseDir,repoDir).value
|
|
|
|
var root: Hash256
|
|
|
|
for w in file.undumpNextStorages:
|
|
|
|
let n = w.seenAccounts - 1 # storages selector based on accounts
|
2022-08-24 13:44:18 +00:00
|
|
|
if n < sample.firstItem:
|
|
|
|
continue
|
|
|
|
if sample.lastItem < n:
|
|
|
|
break
|
|
|
|
if sample.firstItem == n:
|
|
|
|
root = w.root
|
|
|
|
elif w.root != root:
|
|
|
|
break
|
|
|
|
result.add w
|
2022-07-01 11:42:17 +00:00
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
proc to(b: openArray[byte]; T: type ByteArray32): T =
|
|
|
|
## Convert to other representation (or exception)
|
|
|
|
if b.len == 32:
|
|
|
|
(addr result[0]).copyMem(unsafeAddr b[0], 32)
|
|
|
|
else:
|
|
|
|
doAssert b.len == 32
|
|
|
|
|
|
|
|
proc to(b: openArray[byte]; T: type ByteArray33): T =
|
|
|
|
## Convert to other representation (or exception)
|
|
|
|
if b.len == 33:
|
|
|
|
(addr result[0]).copyMem(unsafeAddr b[0], 33)
|
|
|
|
else:
|
|
|
|
doAssert b.len == 33
|
|
|
|
|
|
|
|
proc to(b: ByteArray32|ByteArray33; T: type Blob): T =
|
|
|
|
b.toSeq
|
|
|
|
|
|
|
|
proc to(b: openArray[byte]; T: type NodeTag): T =
|
|
|
|
## Convert from serialised equivalent
|
|
|
|
UInt256.fromBytesBE(b).T
|
|
|
|
|
|
|
|
proc to(w: (byte, NodeTag); T: type Blob): T =
|
|
|
|
let (b,t) = w
|
|
|
|
@[b] & toSeq(t.UInt256.toBytesBE)
|
2022-07-01 11:42:17 +00:00
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
proc to(t: NodeTag; T: type Blob): T =
|
|
|
|
toSeq(t.UInt256.toBytesBE)
|
|
|
|
|
|
|
|
proc flushDbDir(s: string; subDir = "") =
|
2022-07-01 11:42:17 +00:00
|
|
|
if s != "":
|
|
|
|
let baseDir = s / "tmp"
|
2022-08-12 15:42:07 +00:00
|
|
|
for n in 0 ..< nTestDbInstances:
|
|
|
|
let instDir = if subDir == "": baseDir / $n else: baseDir / subDir / $n
|
2022-07-01 11:42:17 +00:00
|
|
|
if (instDir / "nimbus" / "data").dirExists:
|
|
|
|
# Typically under Windows: there might be stale file locks.
|
|
|
|
try: instDir.removeDir except: discard
|
2022-08-12 15:42:07 +00:00
|
|
|
try: (baseDir / subDir).removeDir except: discard
|
2022-07-01 11:42:17 +00:00
|
|
|
block dontClearUnlessEmpty:
|
|
|
|
for w in baseDir.walkDir:
|
|
|
|
break dontClearUnlessEmpty
|
|
|
|
try: baseDir.removeDir except: discard
|
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
proc testDbs(workDir = ""; subDir = ""; instances = nTestDbInstances): TestDbs =
|
2022-07-01 11:42:17 +00:00
|
|
|
if disablePersistentDB or workDir == "":
|
|
|
|
result.persistent = false
|
|
|
|
result.dbDir = "*notused*"
|
|
|
|
else:
|
|
|
|
result.persistent = true
|
2022-08-12 15:42:07 +00:00
|
|
|
if subDir != "":
|
|
|
|
result.dbDir = workDir / "tmp" / subDir
|
|
|
|
else:
|
|
|
|
result.dbDir = workDir / "tmp"
|
2022-07-01 11:42:17 +00:00
|
|
|
if result.persistent:
|
|
|
|
result.dbDir.flushDbDir
|
2022-09-02 18:16:09 +00:00
|
|
|
for n in 0 ..< min(result.cdb.len, instances):
|
2022-08-12 15:42:07 +00:00
|
|
|
result.cdb[n] = (result.dbDir / $n).newChainDB
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
proc lastTwo(a: openArray[string]): seq[string] =
|
|
|
|
if 1 < a.len: @[a[^2],a[^1]] else: a.toSeq
|
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
proc flatten(list: openArray[seq[Blob]]): seq[Blob] =
|
|
|
|
for w in list:
|
|
|
|
result.add w
|
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
proc thisRecord(r: rocksdb_iterator_t): (Blob,Blob) =
|
|
|
|
var kLen, vLen: csize_t
|
|
|
|
let
|
|
|
|
kData = r.rocksdb_iter_key(addr kLen)
|
|
|
|
vData = r.rocksdb_iter_value(addr vLen)
|
|
|
|
if not kData.isNil and not vData.isNil:
|
|
|
|
let
|
|
|
|
key = string.fromBytes(toOpenArrayByte(kData,0,int(kLen)-1))
|
|
|
|
value = string.fromBytes(toOpenArrayByte(vData,0,int(vLen)-1))
|
|
|
|
return (key.mapIt(it.byte),value.mapIt(it.byte))
|
|
|
|
|
|
|
|
proc meanStdDev(sum, sqSum: float; length: int): (float,float) =
|
|
|
|
if 0 < length:
|
|
|
|
result[0] = sum / length.float
|
|
|
|
result[1] = sqrt(sqSum / length.float - result[0] * result[0])
|
|
|
|
|
2022-07-01 11:42:17 +00:00
|
|
|
# ------------------------------------------------------------------------------
|
2022-09-02 18:16:09 +00:00
|
|
|
# Test Runners: accounts and accounts storages
|
2022-07-01 11:42:17 +00:00
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
proc accountsRunner(noisy = true; persistent = true; sample = accSample) =
|
2022-07-01 11:42:17 +00:00
|
|
|
let
|
2022-08-04 08:04:30 +00:00
|
|
|
peer = Peer.new
|
2022-09-02 18:16:09 +00:00
|
|
|
accountsList = sample.to(seq[UndumpAccounts])
|
|
|
|
root = accountsList[0].root
|
2022-08-12 15:42:07 +00:00
|
|
|
tmpDir = getTmpDir()
|
2022-09-02 18:16:09 +00:00
|
|
|
db = if persistent: tmpDir.testDbs(sample.name, instances=2) else: testDbs()
|
2022-07-01 11:42:17 +00:00
|
|
|
dbDir = db.dbDir.split($DirSep).lastTwo.join($DirSep)
|
2022-08-15 15:51:50 +00:00
|
|
|
info = if db.persistent: &"persistent db on \"{dbDir}\""
|
|
|
|
else: "in-memory db"
|
2022-09-02 18:16:09 +00:00
|
|
|
fileInfo = sample.file.splitPath.tail.replace(".txt.gz","")
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
defer:
|
|
|
|
if db.persistent:
|
2022-09-02 18:16:09 +00:00
|
|
|
if not db.cdb[0].rocksStoreRef.isNil:
|
|
|
|
db.cdb[0].rocksStoreRef.store.db.rocksdb_close
|
|
|
|
db.cdb[1].rocksStoreRef.store.db.rocksdb_close
|
2022-08-12 15:42:07 +00:00
|
|
|
tmpDir.flushDbDir(sample.name)
|
2022-07-01 11:42:17 +00:00
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
suite &"SyncSnap: {fileInfo} accounts and proofs for {info}":
|
2022-08-15 15:51:50 +00:00
|
|
|
var
|
2022-10-14 16:40:32 +00:00
|
|
|
desc: SnapDbAccountsRef
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
accKeys: seq[NodeKey]
|
2022-11-25 14:56:42 +00:00
|
|
|
accBaseTag: NodeTag
|
2022-08-04 08:04:30 +00:00
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
test &"Snap-proofing {accountsList.len} items for state root ..{root.pp}":
|
2022-09-16 07:24:12 +00:00
|
|
|
let
|
2022-10-08 17:20:50 +00:00
|
|
|
dbBase = if persistent: SnapDbRef.init(db.cdb[0])
|
|
|
|
else: SnapDbRef.init(newMemoryDB())
|
2022-10-14 16:40:32 +00:00
|
|
|
dbDesc = SnapDbAccountsRef.init(dbBase, root, peer)
|
2022-09-02 18:16:09 +00:00
|
|
|
for n,w in accountsList:
|
2022-11-16 23:51:06 +00:00
|
|
|
check dbDesc.importAccounts(w.base, w.data, persistent).isImportOk
|
2022-08-04 08:04:30 +00:00
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
test &"Merging {accountsList.len} proofs for state root ..{root.pp}":
|
2022-10-08 17:20:50 +00:00
|
|
|
let dbBase = if persistent: SnapDbRef.init(db.cdb[1])
|
|
|
|
else: SnapDbRef.init(newMemoryDB())
|
2022-10-14 16:40:32 +00:00
|
|
|
desc = SnapDbAccountsRef.init(dbBase, root, peer)
|
2022-08-17 07:30:11 +00:00
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
# Load/accumulate data from several samples (needs some particular sort)
|
2022-11-25 14:56:42 +00:00
|
|
|
accBaseTag = accountsList.mapIt(it.base).sortMerge
|
|
|
|
let packed = PackedAccountRange(
|
|
|
|
accounts: accountsList.mapIt(it.data.accounts).sortMerge,
|
|
|
|
proof: accountsList.mapIt(it.data.proof).flatten)
|
2022-11-08 18:56:04 +00:00
|
|
|
# Merging intervals will produce gaps, so the result is expected OK but
|
2022-11-16 23:51:06 +00:00
|
|
|
# different from `.isImportOk`
|
2022-11-25 14:56:42 +00:00
|
|
|
check desc.importAccounts(accBaseTag, packed, true).isOk
|
2022-08-17 07:30:11 +00:00
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
# check desc.merge(lowerBound, accounts) == OkHexDb
|
2022-08-17 07:30:11 +00:00
|
|
|
desc.assignPrettyKeys() # for debugging, make sure that state root ~ "$0"
|
|
|
|
|
2022-08-24 13:44:18 +00:00
|
|
|
# Update list of accounts. There might be additional accounts in the set
|
|
|
|
# of proof nodes, typically before the `lowerBound` of each block. As
|
|
|
|
# there is a list of account ranges (that were merged for testing), one
|
|
|
|
# need to check for additional records only on either end of a range.
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
var keySet = packed.accounts.mapIt(it.accKey).toHashSet
|
2022-09-02 18:16:09 +00:00
|
|
|
for w in accountsList:
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
var key = desc.prevAccountsChainDbKey(w.data.accounts[0].accKey)
|
2022-08-24 13:44:18 +00:00
|
|
|
while key.isOk and key.value notin keySet:
|
|
|
|
keySet.incl key.value
|
2022-10-19 10:04:06 +00:00
|
|
|
let newKey = desc.prevAccountsChainDbKey(key.value)
|
2022-08-24 13:44:18 +00:00
|
|
|
check newKey != key
|
|
|
|
key = newKey
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
key = desc.nextAccountsChainDbKey(w.data.accounts[^1].accKey)
|
2022-08-24 13:44:18 +00:00
|
|
|
while key.isOk and key.value notin keySet:
|
|
|
|
keySet.incl key.value
|
2022-10-19 10:04:06 +00:00
|
|
|
let newKey = desc.nextAccountsChainDbKey(key.value)
|
2022-08-24 13:44:18 +00:00
|
|
|
check newKey != key
|
|
|
|
key = newKey
|
|
|
|
accKeys = toSeq(keySet).mapIt(it.to(NodeTag)).sorted(cmp)
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
.mapIt(it.to(NodeKey))
|
2022-09-02 18:16:09 +00:00
|
|
|
check packed.accounts.len <= accKeys.len
|
2022-08-24 13:44:18 +00:00
|
|
|
|
|
|
|
test &"Revisiting {accKeys.len} items stored items on BaseChainDb":
|
|
|
|
var
|
|
|
|
nextAccount = accKeys[0]
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
prevAccount: NodeKey
|
2022-08-24 13:44:18 +00:00
|
|
|
count = 0
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
for accKey in accKeys:
|
2022-08-24 13:44:18 +00:00
|
|
|
count.inc
|
2022-08-15 15:51:50 +00:00
|
|
|
let
|
2022-08-24 13:44:18 +00:00
|
|
|
pfx = $count & "#"
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
byChainDB = desc.getAccountsChainDb(accKey)
|
|
|
|
byNextKey = desc.nextAccountsChainDbKey(accKey)
|
|
|
|
byPrevKey = desc.prevAccountsChainDbKey(accKey)
|
2022-08-15 15:51:50 +00:00
|
|
|
noisy.say "*** find",
|
2022-09-02 18:16:09 +00:00
|
|
|
"<", count, "> byChainDb=", byChainDB.pp
|
2022-08-15 15:51:50 +00:00
|
|
|
check byChainDB.isOk
|
2022-08-24 13:44:18 +00:00
|
|
|
|
|
|
|
# Check `next` traversal funcionality. If `byNextKey.isOk` fails, the
|
|
|
|
# `nextAccount` value is still the old one and will be different from
|
|
|
|
# the account in the next for-loop cycle (if any.)
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
check pfx & accKey.pp(false) == pfx & nextAccount.pp(false)
|
2022-08-24 13:44:18 +00:00
|
|
|
if byNextKey.isOk:
|
|
|
|
nextAccount = byNextKey.value
|
|
|
|
else:
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
nextAccount = NodeKey.default
|
2022-08-24 13:44:18 +00:00
|
|
|
|
|
|
|
# Check `prev` traversal funcionality
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
if prevAccount != NodeKey.default:
|
2022-08-24 13:44:18 +00:00
|
|
|
check byPrevKey.isOk
|
|
|
|
if byPrevKey.isOk:
|
|
|
|
check pfx & byPrevKey.value.pp(false) == pfx & prevAccount.pp(false)
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
prevAccount = accKey
|
2022-08-24 13:44:18 +00:00
|
|
|
|
2022-08-17 07:30:11 +00:00
|
|
|
# Hexary trie memory database dump. These are key value pairs for
|
|
|
|
# ::
|
|
|
|
# Branch: ($1,b(<$2,$3,..,$17>,))
|
|
|
|
# Extension: ($18,e(832b5e..06e697,$19))
|
|
|
|
# Leaf: ($20,l(cc9b5d..1c3b4,f84401..f9e5129d[#70]))
|
|
|
|
#
|
|
|
|
# where keys are typically represented as `$<id>` or `¶<id>` or `ø`
|
|
|
|
# depending on whether a key is final (`$<id>`), temporary (`¶<id>`)
|
|
|
|
# or unset/missing (`ø`).
|
|
|
|
#
|
|
|
|
# The node types are indicated by a letter after the first key before
|
|
|
|
# the round brackets
|
|
|
|
# ::
|
|
|
|
# Branch: 'b', 'þ', or 'B'
|
|
|
|
# Extension: 'e', '€', or 'E'
|
|
|
|
# Leaf: 'l', 'ł', or 'L'
|
|
|
|
#
|
|
|
|
# Here a small letter indicates a `Static` node which was from the
|
|
|
|
# original `proofs` list, a capital letter indicates a `Mutable` node
|
|
|
|
# added on the fly which might need some change, and the decorated
|
|
|
|
# letters stand for `Locked` nodes which are like `Static` ones but
|
|
|
|
# added later (typically these nodes are update `Mutable` nodes.)
|
|
|
|
#
|
2022-08-24 13:44:18 +00:00
|
|
|
# Beware: dumping a large database is not recommended
|
2022-11-25 14:56:42 +00:00
|
|
|
#true.say "***", "database dump\n ", desc.dumpHexaDB()
|
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
|
2022-11-16 23:51:06 +00:00
|
|
|
test &"Storing/retrieving {accKeys.len} items " &
|
|
|
|
"on persistent state root registry":
|
|
|
|
if not persistent:
|
|
|
|
skip()
|
|
|
|
else:
|
|
|
|
let
|
|
|
|
dbBase = SnapDbRef.init(db.cdb[0])
|
2022-11-25 14:56:42 +00:00
|
|
|
processed = @[(1.to(NodeTag),2.to(NodeTag)),
|
|
|
|
(4.to(NodeTag),5.to(NodeTag)),
|
|
|
|
(6.to(NodeTag),7.to(NodeTag))]
|
2022-11-16 23:51:06 +00:00
|
|
|
slotAccounts = seq[NodeKey].default
|
|
|
|
for n,w in accKeys:
|
|
|
|
check dbBase.savePivot(
|
2022-11-25 14:56:42 +00:00
|
|
|
SnapDbPivotRegistry(
|
|
|
|
header: BlockHeader(stateRoot: w.to(Hash256)),
|
|
|
|
nAccounts: n.uint64,
|
|
|
|
nSlotLists: n.uint64,
|
|
|
|
processed: processed,
|
|
|
|
slotAccounts: slotAccounts)).isOk
|
2022-11-16 23:51:06 +00:00
|
|
|
# verify latest state root
|
|
|
|
block:
|
|
|
|
let rc = dbBase.recoverPivot()
|
|
|
|
check rc.isOk
|
|
|
|
if rc.isOk:
|
|
|
|
check rc.value.nAccounts == n.uint64
|
|
|
|
check rc.value.nSlotLists == n.uint64
|
2022-11-25 14:56:42 +00:00
|
|
|
check rc.value.processed == processed
|
2022-11-16 23:51:06 +00:00
|
|
|
for n,w in accKeys:
|
|
|
|
block:
|
|
|
|
let rc = dbBase.recoverPivot(w)
|
|
|
|
check rc.isOk
|
|
|
|
if rc.isOk:
|
|
|
|
check rc.value.nAccounts == n.uint64
|
|
|
|
check rc.value.nSlotLists == n.uint64
|
|
|
|
# Update record in place
|
|
|
|
check dbBase.savePivot(
|
2022-11-25 14:56:42 +00:00
|
|
|
SnapDbPivotRegistry(
|
|
|
|
header: BlockHeader(stateRoot: w.to(Hash256)),
|
|
|
|
nAccounts: n.uint64,
|
|
|
|
nSlotLists: 0,
|
|
|
|
processed: @[],
|
|
|
|
slotAccounts: @[])).isOk
|
2022-11-16 23:51:06 +00:00
|
|
|
block:
|
|
|
|
let rc = dbBase.recoverPivot(w)
|
|
|
|
check rc.isOk
|
|
|
|
if rc.isOk:
|
|
|
|
check rc.value.nAccounts == n.uint64
|
|
|
|
check rc.value.nSlotLists == 0
|
2022-11-25 14:56:42 +00:00
|
|
|
check rc.value.processed == seq[(NodeTag,NodeTag)].default
|
2022-08-12 15:42:07 +00:00
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
proc storagesRunner(
|
|
|
|
noisy = true;
|
|
|
|
persistent = true;
|
|
|
|
sample = storSample;
|
|
|
|
knownFailures: seq[(string,seq[(int,HexaryDbError)])] = @[]) =
|
|
|
|
let
|
|
|
|
peer = Peer.new
|
|
|
|
accountsList = sample.to(seq[UndumpAccounts])
|
|
|
|
storagesList = sample.to(seq[UndumpStorages])
|
|
|
|
root = accountsList[0].root
|
|
|
|
tmpDir = getTmpDir()
|
|
|
|
db = if persistent: tmpDir.testDbs(sample.name, instances=1) else: testDbs()
|
|
|
|
dbDir = db.dbDir.split($DirSep).lastTwo.join($DirSep)
|
|
|
|
info = if db.persistent: &"persistent db on \"{dbDir}\""
|
|
|
|
else: "in-memory db"
|
|
|
|
fileInfo = sample.file.splitPath.tail.replace(".txt.gz","")
|
|
|
|
|
|
|
|
defer:
|
|
|
|
if db.persistent:
|
|
|
|
if not db.cdb[0].rocksStoreRef.isNil:
|
|
|
|
db.cdb[0].rocksStoreRef.store.db.rocksdb_close
|
|
|
|
tmpDir.flushDbDir(sample.name)
|
|
|
|
|
|
|
|
suite &"SyncSnap: {fileInfo} accounts storage for {info}":
|
|
|
|
let
|
2022-10-08 17:20:50 +00:00
|
|
|
dbBase = if persistent: SnapDbRef.init(db.cdb[0])
|
|
|
|
else: SnapDbRef.init(newMemoryDB())
|
2022-09-02 18:16:09 +00:00
|
|
|
|
|
|
|
test &"Merging {accountsList.len} accounts for state root ..{root.pp}":
|
|
|
|
for w in accountsList:
|
2022-10-14 16:40:32 +00:00
|
|
|
let desc = SnapDbAccountsRef.init(dbBase, root, peer)
|
2022-11-16 23:51:06 +00:00
|
|
|
check desc.importAccounts(w.base, w.data, persistent).isImportOk
|
2022-09-02 18:16:09 +00:00
|
|
|
|
|
|
|
test &"Merging {storagesList.len} storages lists":
|
2022-09-16 07:24:12 +00:00
|
|
|
let
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
dbDesc = SnapDbStorageSlotsRef.init(
|
|
|
|
dbBase, Hash256().to(NodeKey), Hash256(), peer)
|
2022-09-16 07:24:12 +00:00
|
|
|
ignore = knownFailures.toTable
|
2022-09-02 18:16:09 +00:00
|
|
|
for n,w in storagesList:
|
|
|
|
let
|
|
|
|
testId = fileInfo & "#" & $n
|
|
|
|
expRc = if ignore.hasKey(testId):
|
|
|
|
Result[void,seq[(int,HexaryDbError)]].err(ignore[testId])
|
|
|
|
else:
|
|
|
|
OkStoDb
|
2022-10-19 10:04:06 +00:00
|
|
|
check dbDesc.importStorageSlots(w.data, persistent).toStoDbRc == expRc
|
2022-09-02 18:16:09 +00:00
|
|
|
|
2022-10-20 16:59:54 +00:00
|
|
|
test &"Inspecting {storagesList.len} imported storages lists sub-tries":
|
|
|
|
let ignore = knownFailures.toTable
|
|
|
|
for n,w in storagesList:
|
|
|
|
let
|
|
|
|
testId = fileInfo & "#" & $n
|
|
|
|
errInx = if ignore.hasKey(testId): ignore[testId][0][0]
|
|
|
|
else: high(int)
|
|
|
|
for m in 0 ..< w.data.storages.len:
|
|
|
|
let
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
accKey = w.data.storages[m].account.accKey
|
2022-10-20 16:59:54 +00:00
|
|
|
root = w.data.storages[m].account.storageRoot
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
dbDesc = SnapDbStorageSlotsRef.init(dbBase, accKey, root, peer)
|
2022-10-20 16:59:54 +00:00
|
|
|
rc = dbDesc.inspectStorageSlotsTrie(persistent=persistent)
|
|
|
|
if m == errInx:
|
|
|
|
check rc == Result[TrieNodeStat,HexaryDbError].err(TrieIsEmpty)
|
|
|
|
else:
|
|
|
|
check rc.isOk # ok => level > 0 and not stopped
|
|
|
|
|
|
|
|
|
2022-09-16 07:24:12 +00:00
|
|
|
proc inspectionRunner(
|
|
|
|
noisy = true;
|
|
|
|
persistent = true;
|
|
|
|
cascaded = true;
|
|
|
|
sample: openArray[AccountsSample] = snapTestList) =
|
|
|
|
let
|
|
|
|
peer = Peer.new
|
|
|
|
inspectList = sample.mapIt(it.to(seq[UndumpAccounts]))
|
|
|
|
tmpDir = getTmpDir()
|
|
|
|
db = if persistent: tmpDir.testDbs(sample[0].name) else: testDbs()
|
|
|
|
dbDir = db.dbDir.split($DirSep).lastTwo.join($DirSep)
|
|
|
|
info = if db.persistent: &"persistent db on \"{dbDir}\""
|
|
|
|
else: "in-memory db"
|
|
|
|
fileInfo = "[" & sample[0].file.splitPath.tail.replace(".txt.gz","") & "..]"
|
|
|
|
|
|
|
|
defer:
|
|
|
|
if db.persistent:
|
|
|
|
for n in 0 ..< nTestDbInstances:
|
|
|
|
if db.cdb[n].rocksStoreRef.isNil:
|
|
|
|
break
|
|
|
|
db.cdb[n].rocksStoreRef.store.db.rocksdb_close
|
|
|
|
tmpDir.flushDbDir(sample[0].name)
|
|
|
|
|
|
|
|
suite &"SyncSnap: inspect {fileInfo} lists for {info} for healing":
|
|
|
|
let
|
2022-10-08 17:20:50 +00:00
|
|
|
memBase = SnapDbRef.init(newMemoryDB())
|
2022-10-14 16:40:32 +00:00
|
|
|
memDesc = SnapDbAccountsRef.init(memBase, Hash256(), peer)
|
2022-09-16 07:24:12 +00:00
|
|
|
var
|
|
|
|
singleStats: seq[(int,TrieNodeStat)]
|
|
|
|
accuStats: seq[(int,TrieNodeStat)]
|
2022-10-08 17:20:50 +00:00
|
|
|
perBase,altBase: SnapDbRef
|
2022-10-14 16:40:32 +00:00
|
|
|
perDesc,altDesc: SnapDbAccountsRef
|
2022-09-16 07:24:12 +00:00
|
|
|
if persistent:
|
2022-10-08 17:20:50 +00:00
|
|
|
perBase = SnapDbRef.init(db.cdb[0])
|
2022-10-14 16:40:32 +00:00
|
|
|
perDesc = SnapDbAccountsRef.init(perBase, Hash256(), peer)
|
2022-10-08 17:20:50 +00:00
|
|
|
altBase = SnapDbRef.init(db.cdb[1])
|
2022-10-14 16:40:32 +00:00
|
|
|
altDesc = SnapDbAccountsRef.init(altBase, Hash256(), peer)
|
2022-09-16 07:24:12 +00:00
|
|
|
|
|
|
|
test &"Fingerprinting {inspectList.len} single accounts lists " &
|
|
|
|
"for in-memory-db":
|
|
|
|
for n,accList in inspectList:
|
|
|
|
# Separate storage
|
|
|
|
let
|
|
|
|
root = accList[0].root
|
|
|
|
rootKey = root.to(NodeKey)
|
2022-10-14 16:40:32 +00:00
|
|
|
desc = SnapDbAccountsRef.init(memBase, root, peer)
|
2022-09-16 07:24:12 +00:00
|
|
|
for w in accList:
|
2022-11-16 23:51:06 +00:00
|
|
|
check desc.importAccounts(w.base, w.data, persistent=false).isImportOk
|
2022-09-16 07:24:12 +00:00
|
|
|
let rc = desc.inspectAccountsTrie(persistent=false)
|
|
|
|
check rc.isOk
|
|
|
|
let
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
dangling = rc.value.dangling.mapIt(it.partialPath)
|
2022-11-16 23:51:06 +00:00
|
|
|
keys = desc.hexaDb.hexaryInspectToKeys(rootKey, dangling)
|
2022-09-16 07:24:12 +00:00
|
|
|
check dangling.len == keys.len
|
2022-10-14 16:40:32 +00:00
|
|
|
singleStats.add (desc.hexaDb.tab.len,rc.value)
|
2022-09-16 07:24:12 +00:00
|
|
|
|
2022-11-16 23:51:06 +00:00
|
|
|
# Verify piecemeal approach for `inspectAccountsTrie()` ...
|
|
|
|
var
|
|
|
|
ctx = TrieNodeStatCtxRef()
|
|
|
|
piecemeal: HashSet[Blob]
|
|
|
|
while not ctx.isNil:
|
|
|
|
let rx = desc.inspectAccountsTrie(
|
|
|
|
resumeCtx=ctx, suspendAfter=128, persistent=false)
|
|
|
|
check rx.isOk
|
|
|
|
let stats = rx.get(otherwise = TrieNodeStat())
|
|
|
|
ctx = stats.resumeCtx
|
|
|
|
piecemeal.incl stats.dangling.mapIt(it.partialPath).toHashSet
|
|
|
|
# Must match earlier all-in-one result
|
|
|
|
check dangling.len == piecemeal.len
|
|
|
|
check dangling.toHashSet == piecemeal
|
|
|
|
|
2022-09-16 07:24:12 +00:00
|
|
|
test &"Fingerprinting {inspectList.len} single accounts lists " &
|
|
|
|
"for persistent db":
|
|
|
|
if not persistent:
|
|
|
|
skip()
|
|
|
|
else:
|
|
|
|
for n,accList in inspectList:
|
|
|
|
if nTestDbInstances <= 2+n or db.cdb[2+n].rocksStoreRef.isNil:
|
|
|
|
continue
|
|
|
|
# Separate storage on persistent DB (leaving first db slot empty)
|
|
|
|
let
|
|
|
|
root = accList[0].root
|
|
|
|
rootKey = root.to(NodeKey)
|
2022-10-08 17:20:50 +00:00
|
|
|
dbBase = SnapDbRef.init(db.cdb[2+n])
|
2022-10-14 16:40:32 +00:00
|
|
|
desc = SnapDbAccountsRef.init(dbBase, root, peer)
|
2022-11-25 14:56:42 +00:00
|
|
|
|
2022-09-16 07:24:12 +00:00
|
|
|
for w in accList:
|
2022-11-16 23:51:06 +00:00
|
|
|
check desc.importAccounts(w.base,w.data, persistent=true).isImportOk
|
|
|
|
let rc = desc.inspectAccountsTrie(persistent=true)
|
2022-09-16 07:24:12 +00:00
|
|
|
check rc.isOk
|
|
|
|
let
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
dangling = rc.value.dangling.mapIt(it.partialPath)
|
2022-11-16 23:51:06 +00:00
|
|
|
keys = desc.hexaDb.hexaryInspectToKeys(rootKey, dangling)
|
2022-09-16 07:24:12 +00:00
|
|
|
check dangling.len == keys.len
|
|
|
|
# Must be the same as the in-memory fingerprint
|
2022-11-16 23:51:06 +00:00
|
|
|
let ssn1 = singleStats[n][1].dangling.mapIt(it.partialPath)
|
|
|
|
check ssn1.toHashSet == dangling.toHashSet
|
|
|
|
|
|
|
|
# Verify piecemeal approach for `inspectAccountsTrie()` ...
|
|
|
|
var
|
|
|
|
ctx = TrieNodeStatCtxRef()
|
|
|
|
piecemeal: HashSet[Blob]
|
|
|
|
while not ctx.isNil:
|
|
|
|
let rx = desc.inspectAccountsTrie(
|
|
|
|
resumeCtx=ctx, suspendAfter=128, persistent=persistent)
|
|
|
|
check rx.isOk
|
|
|
|
let stats = rx.get(otherwise = TrieNodeStat())
|
|
|
|
ctx = stats.resumeCtx
|
|
|
|
piecemeal.incl stats.dangling.mapIt(it.partialPath).toHashSet
|
|
|
|
# Must match earlier all-in-one result
|
|
|
|
check dangling.len == piecemeal.len
|
|
|
|
check dangling.toHashSet == piecemeal
|
2022-09-16 07:24:12 +00:00
|
|
|
|
|
|
|
test &"Fingerprinting {inspectList.len} accumulated accounts lists " &
|
|
|
|
"for in-memory-db":
|
|
|
|
for n,accList in inspectList:
|
|
|
|
# Accumulated storage
|
|
|
|
let
|
|
|
|
root = accList[0].root
|
|
|
|
rootKey = root.to(NodeKey)
|
|
|
|
desc = memDesc.dup(root,Peer())
|
|
|
|
for w in accList:
|
2022-11-16 23:51:06 +00:00
|
|
|
check desc.importAccounts(w.base, w.data, persistent=false).isImportOk
|
2022-09-16 07:24:12 +00:00
|
|
|
let rc = desc.inspectAccountsTrie(persistent=false)
|
|
|
|
check rc.isOk
|
|
|
|
let
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
dangling = rc.value.dangling.mapIt(it.partialPath)
|
2022-10-14 16:40:32 +00:00
|
|
|
keys = desc.hexaDb.hexaryInspectToKeys(
|
2022-09-16 07:24:12 +00:00
|
|
|
rootKey, dangling.toHashSet.toSeq)
|
|
|
|
check dangling.len == keys.len
|
2022-10-14 16:40:32 +00:00
|
|
|
accuStats.add (desc.hexaDb.tab.len,rc.value)
|
2022-09-16 07:24:12 +00:00
|
|
|
|
|
|
|
test &"Fingerprinting {inspectList.len} accumulated accounts lists " &
|
|
|
|
"for persistent db":
|
|
|
|
if not persistent:
|
|
|
|
skip()
|
|
|
|
else:
|
|
|
|
for n,accList in inspectList:
|
|
|
|
# Accumulated storage on persistent DB (using first db slot)
|
|
|
|
let
|
|
|
|
root = accList[0].root
|
|
|
|
rootKey = root.to(NodeKey)
|
|
|
|
rootSet = [rootKey].toHashSet
|
|
|
|
desc = perDesc.dup(root,Peer())
|
|
|
|
for w in accList:
|
2022-11-16 23:51:06 +00:00
|
|
|
check desc.importAccounts(w.base, w.data, persistent).isImportOk
|
2022-09-16 07:24:12 +00:00
|
|
|
let rc = desc.inspectAccountsTrie(persistent=false)
|
|
|
|
check rc.isOk
|
|
|
|
let
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
dangling = rc.value.dangling.mapIt(it.partialPath)
|
2022-10-14 16:40:32 +00:00
|
|
|
keys = desc.hexaDb.hexaryInspectToKeys(
|
2022-09-16 07:24:12 +00:00
|
|
|
rootKey, dangling.toHashSet.toSeq)
|
|
|
|
check dangling.len == keys.len
|
|
|
|
check accuStats[n][1] == rc.value
|
|
|
|
|
|
|
|
test &"Cascaded fingerprinting {inspectList.len} accumulated accounts " &
|
|
|
|
"lists for in-memory-db":
|
|
|
|
if not cascaded:
|
|
|
|
skip()
|
|
|
|
else:
|
|
|
|
let
|
2022-10-08 17:20:50 +00:00
|
|
|
cscBase = SnapDbRef.init(newMemoryDB())
|
2022-10-14 16:40:32 +00:00
|
|
|
cscDesc = SnapDbAccountsRef.init(cscBase, Hash256(), peer)
|
2022-09-16 07:24:12 +00:00
|
|
|
var
|
|
|
|
cscStep: Table[NodeKey,(int,seq[Blob])]
|
|
|
|
for n,accList in inspectList:
|
|
|
|
# Accumulated storage
|
|
|
|
let
|
|
|
|
root = accList[0].root
|
|
|
|
rootKey = root.to(NodeKey)
|
|
|
|
desc = cscDesc.dup(root,Peer())
|
|
|
|
for w in accList:
|
2022-11-16 23:51:06 +00:00
|
|
|
check desc.importAccounts(w.base,w.data,persistent=false).isImportOk
|
2022-09-16 07:24:12 +00:00
|
|
|
if cscStep.hasKeyOrPut(rootKey,(1,seq[Blob].default)):
|
|
|
|
cscStep[rootKey][0].inc
|
|
|
|
let
|
|
|
|
r0 = desc.inspectAccountsTrie(persistent=false)
|
2022-10-08 17:20:50 +00:00
|
|
|
rc = desc.inspectAccountsTrie(cscStep[rootKey][1],persistent=false)
|
2022-09-16 07:24:12 +00:00
|
|
|
check rc.isOk
|
|
|
|
let
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
accumulated = r0.value.dangling.mapIt(it.partialPath).toHashSet
|
|
|
|
cascaded = rc.value.dangling.mapIt(it.partialPath).toHashSet
|
2022-09-16 07:24:12 +00:00
|
|
|
check accumulated == cascaded
|
|
|
|
# Make sure that there are no trivial cases
|
|
|
|
let trivialCases = toSeq(cscStep.values).filterIt(it[0] <= 1).len
|
|
|
|
check trivialCases == 0
|
|
|
|
|
|
|
|
test &"Cascaded fingerprinting {inspectList.len} accumulated accounts " &
|
|
|
|
"for persistent db":
|
|
|
|
if not cascaded or not persistent:
|
|
|
|
skip()
|
|
|
|
else:
|
|
|
|
let
|
|
|
|
cscBase = altBase
|
|
|
|
cscDesc = altDesc
|
|
|
|
var
|
|
|
|
cscStep: Table[NodeKey,(int,seq[Blob])]
|
|
|
|
for n,accList in inspectList:
|
|
|
|
# Accumulated storage
|
|
|
|
let
|
|
|
|
root = accList[0].root
|
|
|
|
rootKey = root.to(NodeKey)
|
|
|
|
desc = cscDesc.dup(root,Peer())
|
|
|
|
for w in accList:
|
2022-11-16 23:51:06 +00:00
|
|
|
check desc.importAccounts(w.base,w.data, persistent).isImportOk
|
2022-09-16 07:24:12 +00:00
|
|
|
if cscStep.hasKeyOrPut(rootKey,(1,seq[Blob].default)):
|
|
|
|
cscStep[rootKey][0].inc
|
|
|
|
let
|
|
|
|
r0 = desc.inspectAccountsTrie(persistent=true)
|
2022-10-08 17:20:50 +00:00
|
|
|
rc = desc.inspectAccountsTrie(cscStep[rootKey][1],persistent=true)
|
2022-09-16 07:24:12 +00:00
|
|
|
check rc.isOk
|
|
|
|
let
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
accumulated = r0.value.dangling.mapIt(it.partialPath).toHashSet
|
|
|
|
cascaded = rc.value.dangling.mapIt(it.partialPath).toHashSet
|
2022-09-16 07:24:12 +00:00
|
|
|
check accumulated == cascaded
|
|
|
|
# Make sure that there are no trivial cases
|
|
|
|
let trivialCases = toSeq(cscStep.values).filterIt(it[0] <= 1).len
|
|
|
|
check trivialCases == 0
|
2022-09-02 18:16:09 +00:00
|
|
|
|
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
# Test Runners: database timing tests
|
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
|
|
|
|
proc importRunner(noisy = true; persistent = true; capture = bChainCapture) =
|
2022-08-12 15:42:07 +00:00
|
|
|
|
|
|
|
let
|
|
|
|
fileInfo = capture.file.splitFile.name.split(".")[0]
|
|
|
|
filePath = capture.file.findFilePath(baseDir,repoDir).value
|
|
|
|
tmpDir = getTmpDir()
|
|
|
|
db = if persistent: tmpDir.testDbs(capture.name) else: testDbs()
|
2022-08-15 15:51:50 +00:00
|
|
|
numBlocksInfo = if capture.numBlocks == high(int): ""
|
|
|
|
else: $capture.numBlocks & " "
|
2022-08-12 15:42:07 +00:00
|
|
|
loadNoise = noisy
|
|
|
|
|
|
|
|
defer:
|
|
|
|
if db.persistent:
|
|
|
|
tmpDir.flushDbDir(capture.name)
|
2022-08-04 08:04:30 +00:00
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
suite &"SyncSnap: using {fileInfo} capture for testing db timings":
|
|
|
|
var
|
|
|
|
ddb: BaseChainDB # perstent DB on disk
|
|
|
|
chn: Chain
|
|
|
|
|
|
|
|
test &"Create persistent BaseChainDB on {tmpDir}":
|
|
|
|
let chainDb = if db.persistent: db.cdb[0].trieDB
|
|
|
|
else: newMemoryDB()
|
|
|
|
|
|
|
|
# Constructor ...
|
|
|
|
ddb = newBaseChainDB(
|
|
|
|
chainDb,
|
|
|
|
id = capture.network,
|
|
|
|
pruneTrie = true,
|
|
|
|
params = capture.network.networkParams)
|
|
|
|
|
|
|
|
ddb.initializeEmptyDb
|
|
|
|
chn = ddb.newChain
|
|
|
|
|
|
|
|
test &"Storing {numBlocksInfo}persistent blocks from dump":
|
|
|
|
for w in filePath.undumpNextGroup:
|
|
|
|
let (fromBlock, toBlock) = (w[0][0].blockNumber, w[0][^1].blockNumber)
|
|
|
|
if fromBlock == 0.u256:
|
|
|
|
doAssert w[0][0] == ddb.getBlockHeader(0.u256)
|
|
|
|
continue
|
|
|
|
# Message if [fromBlock,toBlock] contains a multiple of 700
|
|
|
|
if fromBlock + (toBlock mod 900) <= toBlock:
|
|
|
|
loadNoise.say "***", &"processing ...[#{fromBlock},#{toBlock}]..."
|
|
|
|
check chn.persistBlocks(w[0], w[1]).isOk
|
|
|
|
if capture.numBlocks.toBlockNumber <= w[0][^1].blockNumber:
|
|
|
|
break
|
|
|
|
|
|
|
|
test "Extract key-value records into memory tables via rocksdb iterator":
|
|
|
|
# Implicit test: if not persistent => db.cdb[0] is nil
|
|
|
|
if db.cdb[0].rocksStoreRef.isNil:
|
|
|
|
skip()
|
|
|
|
else:
|
|
|
|
let
|
|
|
|
rdb = db.cdb[0].rocksStoreRef
|
|
|
|
rop = rdb.store.readOptions
|
|
|
|
rit = rdb.store.db.rocksdb_create_iterator(rop)
|
|
|
|
check not rit.isNil
|
|
|
|
|
|
|
|
xTab32.clear
|
|
|
|
xTab33.clear
|
|
|
|
|
|
|
|
rit.rocksdb_iter_seek_to_first()
|
|
|
|
while rit.rocksdb_iter_valid() != 0:
|
|
|
|
let (key,val) = rit.thisRecord()
|
|
|
|
rit.rocksdb_iter_next()
|
|
|
|
if key.len == 32:
|
|
|
|
xTab32[key.to(ByteArray32)] = val
|
|
|
|
xVal32Sum += val.len.float
|
|
|
|
xVal32SqSum += val.len.float * val.len.float
|
|
|
|
check key.to(ByteArray32).to(Blob) == key
|
|
|
|
elif key.len == 33:
|
|
|
|
xTab33[key.to(ByteArray33)] = val
|
|
|
|
xVal33Sum += val.len.float
|
|
|
|
xVal33SqSum += val.len.float * val.len.float
|
|
|
|
check key.to(ByteArray33).to(Blob) == key
|
|
|
|
else:
|
|
|
|
noisy.say "***", "ignoring key=", key.toHex
|
|
|
|
|
|
|
|
rit.rocksdb_iter_destroy()
|
|
|
|
|
|
|
|
var
|
|
|
|
(mean32, stdv32) = meanStdDev(xVal32Sum, xVal32SqSum, xTab32.len)
|
|
|
|
(mean33, stdv33) = meanStdDev(xVal33Sum, xVal33SqSum, xTab33.len)
|
|
|
|
noisy.say "***",
|
|
|
|
"key 32 table: ",
|
|
|
|
&"size={xTab32.len} valLen={(mean32+0.5).int}({(stdv32+0.5).int})",
|
|
|
|
", key 33 table: ",
|
|
|
|
&"size={xTab33.len} valLen={(mean33+0.5).int}({(stdv33+0.5).int})"
|
|
|
|
|
|
|
|
|
|
|
|
proc storeRunner(noisy = true; persistent = true; cleanUp = true) =
|
|
|
|
let
|
|
|
|
fullNoise = false
|
|
|
|
var
|
|
|
|
emptyDb = "empty"
|
|
|
|
|
|
|
|
# Allows to repeat storing on existing data
|
|
|
|
if not xDbs.cdb[0].isNil:
|
|
|
|
emptyDb = "pre-loaded"
|
|
|
|
elif persistent:
|
|
|
|
xTmpDir = getTmpDir()
|
|
|
|
xDbs = xTmpDir.testDbs("store-runner")
|
|
|
|
else:
|
|
|
|
xDbs = testDbs()
|
|
|
|
|
|
|
|
defer:
|
|
|
|
if xDbs.persistent and cleanUp:
|
2022-09-16 07:24:12 +00:00
|
|
|
for n in 0 ..< nTestDbInstances:
|
|
|
|
if xDbs.cdb[n].rocksStoreRef.isNil:
|
|
|
|
break
|
|
|
|
xDbs.cdb[n].rocksStoreRef.store.db.rocksdb_close
|
2022-08-12 15:42:07 +00:00
|
|
|
xTmpDir.flushDbDir("store-runner")
|
|
|
|
xDbs.reset
|
|
|
|
|
|
|
|
suite &"SyncSnap: storage tests on {emptyDb} databases":
|
|
|
|
#
|
|
|
|
# `xDbs` instance slots layout:
|
|
|
|
#
|
|
|
|
# * cdb[0] -- direct db, key length 32, no transaction
|
|
|
|
# * cdb[1] -- direct db, key length 32 as 33, no transaction
|
|
|
|
#
|
|
|
|
# * cdb[2] -- direct db, key length 32, transaction based
|
|
|
|
# * cdb[3] -- direct db, key length 32 as 33, transaction based
|
|
|
|
#
|
|
|
|
# * cdb[4] -- direct db, key length 33, no transaction
|
|
|
|
# * cdb[5] -- direct db, key length 33, transaction based
|
|
|
|
#
|
|
|
|
# * cdb[6] -- rocksdb, key length 32
|
|
|
|
# * cdb[7] -- rocksdb, key length 32 as 33
|
|
|
|
# * cdb[8] -- rocksdb, key length 33
|
|
|
|
#
|
|
|
|
doAssert 9 <= nTestDbInstances
|
2022-09-02 18:16:09 +00:00
|
|
|
doAssert not xDbs.cdb[8].isNil
|
2022-08-12 15:42:07 +00:00
|
|
|
|
|
|
|
if xTab32.len == 0 or xTab33.len == 0:
|
|
|
|
test &"Both tables with 32 byte keys(size={xTab32.len}), " &
|
|
|
|
&"33 byte keys(size={xTab32.len}) must be non-empty":
|
|
|
|
skip()
|
|
|
|
else:
|
|
|
|
# cdb[0] -- direct db, key length 32, no transaction
|
|
|
|
test &"Directly store {xTab32.len} records " &
|
|
|
|
&"(key length 32) into {emptyDb} trie database":
|
|
|
|
var ela: Duration
|
|
|
|
let tdb = xDbs.cdb[0].trieDB
|
|
|
|
|
|
|
|
if noisy: echo ""
|
|
|
|
noisy.showElapsed("Standard db loader(keyLen 32)", ela):
|
|
|
|
for (key,val) in xTab32.pairs:
|
|
|
|
tdb.put(key, val)
|
|
|
|
|
|
|
|
if ela.inNanoseconds != 0:
|
|
|
|
let
|
|
|
|
elaNs = ela.inNanoseconds.float
|
|
|
|
perRec = ((elaNs / xTab32.len.float) + 0.5).int.initDuration
|
|
|
|
noisy.say "***",
|
|
|
|
"nRecords=", xTab32.len, ", ",
|
|
|
|
"perRecord=", perRec.pp
|
|
|
|
|
|
|
|
# cdb[1] -- direct db, key length 32 as 33, no transaction
|
|
|
|
test &"Directly store {xTab32.len} records " &
|
|
|
|
&"(key length 33) into {emptyDb} trie database":
|
|
|
|
var ela = initDuration()
|
|
|
|
let tdb = xDbs.cdb[1].trieDB
|
|
|
|
|
|
|
|
if noisy: echo ""
|
|
|
|
noisy.showElapsed("Standard db loader(keyLen 32 as 33)", ela):
|
|
|
|
for (key,val) in xTab32.pairs:
|
|
|
|
tdb.put(@[99.byte] & key.toSeq, val)
|
|
|
|
|
|
|
|
if ela.inNanoseconds != 0:
|
|
|
|
let
|
|
|
|
elaNs = ela.inNanoseconds.float
|
|
|
|
perRec = ((elaNs / xTab32.len.float) + 0.5).int.initDuration
|
|
|
|
noisy.say "***",
|
|
|
|
"nRecords=", xTab32.len, ", ",
|
|
|
|
"perRecord=", perRec.pp
|
|
|
|
|
|
|
|
# cdb[2] -- direct db, key length 32, transaction based
|
|
|
|
test &"Transactionally store {xTab32.len} records " &
|
|
|
|
&"(key length 32) into {emptyDb} trie database":
|
|
|
|
var ela: Duration
|
|
|
|
let tdb = xDbs.cdb[2].trieDB
|
|
|
|
|
|
|
|
if noisy: echo ""
|
|
|
|
noisy.showElapsed("Standard db loader(tx,keyLen 32)", ela):
|
|
|
|
let dbTx = tdb.beginTransaction
|
|
|
|
defer: dbTx.commit
|
|
|
|
|
|
|
|
for (key,val) in xTab32.pairs:
|
|
|
|
tdb.put(key, val)
|
|
|
|
|
|
|
|
if ela.inNanoseconds != 0:
|
|
|
|
let
|
|
|
|
elaNs = ela.inNanoseconds.float
|
|
|
|
perRec = ((elaNs / xTab32.len.float) + 0.5).int.initDuration
|
|
|
|
noisy.say "***",
|
|
|
|
"nRecords=", xTab32.len, ", ",
|
|
|
|
"perRecord=", perRec.pp
|
|
|
|
|
|
|
|
# cdb[3] -- direct db, key length 32 as 33, transaction based
|
|
|
|
test &"Transactionally store {xTab32.len} records " &
|
|
|
|
&"(key length 33) into {emptyDb} trie database":
|
|
|
|
var ela: Duration
|
|
|
|
let tdb = xDbs.cdb[3].trieDB
|
|
|
|
|
|
|
|
if noisy: echo ""
|
|
|
|
noisy.showElapsed("Standard db loader(tx,keyLen 32 as 33)", ela):
|
|
|
|
let dbTx = tdb.beginTransaction
|
|
|
|
defer: dbTx.commit
|
|
|
|
|
|
|
|
for (key,val) in xTab32.pairs:
|
|
|
|
tdb.put(@[99.byte] & key.toSeq, val)
|
|
|
|
|
|
|
|
if ela.inNanoseconds != 0:
|
|
|
|
let
|
|
|
|
elaNs = ela.inNanoseconds.float
|
|
|
|
perRec = ((elaNs / xTab32.len.float) + 0.5).int.initDuration
|
|
|
|
noisy.say "***",
|
|
|
|
"nRecords=", xTab32.len, ", ",
|
|
|
|
"perRecord=", perRec.pp
|
|
|
|
|
|
|
|
# cdb[4] -- direct db, key length 33, no transaction
|
|
|
|
test &"Directly store {xTab33.len} records " &
|
|
|
|
&"(key length 33) into {emptyDb} trie database":
|
|
|
|
var ela: Duration
|
|
|
|
let tdb = xDbs.cdb[4].trieDB
|
|
|
|
|
|
|
|
if noisy: echo ""
|
|
|
|
noisy.showElapsed("Standard db loader(keyLen 33)", ela):
|
|
|
|
for (key,val) in xTab33.pairs:
|
|
|
|
tdb.put(key, val)
|
|
|
|
|
|
|
|
if ela.inNanoseconds != 0:
|
|
|
|
let
|
|
|
|
elaNs = ela.inNanoseconds.float
|
|
|
|
perRec = ((elaNs / xTab33.len.float) + 0.5).int.initDuration
|
|
|
|
noisy.say "***",
|
|
|
|
"nRecords=", xTab33.len, ", ",
|
|
|
|
"perRecord=", perRec.pp
|
|
|
|
|
|
|
|
# cdb[5] -- direct db, key length 33, transaction based
|
|
|
|
test &"Transactionally store {xTab33.len} records " &
|
|
|
|
&"(key length 33) into {emptyDb} trie database":
|
|
|
|
var ela: Duration
|
|
|
|
let tdb = xDbs.cdb[5].trieDB
|
|
|
|
|
|
|
|
if noisy: echo ""
|
|
|
|
noisy.showElapsed("Standard db loader(tx,keyLen 33)", ela):
|
|
|
|
let dbTx = tdb.beginTransaction
|
|
|
|
defer: dbTx.commit
|
|
|
|
|
|
|
|
for (key,val) in xTab33.pairs:
|
|
|
|
tdb.put(key, val)
|
|
|
|
|
|
|
|
if ela.inNanoseconds != 0:
|
|
|
|
let
|
|
|
|
elaNs = ela.inNanoseconds.float
|
|
|
|
perRec = ((elaNs / xTab33.len.float) + 0.5).int.initDuration
|
|
|
|
noisy.say "***",
|
|
|
|
"nRecords=", xTab33.len, ", ",
|
|
|
|
"perRecord=", perRec.pp
|
|
|
|
|
|
|
|
if xDbs.cdb[0].rocksStoreRef.isNil:
|
2022-09-02 18:16:09 +00:00
|
|
|
test "The rocksdb interface must be available": skip()
|
2022-08-12 15:42:07 +00:00
|
|
|
else:
|
|
|
|
# cdb[6] -- rocksdb, key length 32
|
|
|
|
test &"Store {xTab32.len} records " &
|
|
|
|
"(key length 32) into empty rocksdb table":
|
|
|
|
var
|
|
|
|
ela: array[4,Duration]
|
|
|
|
size: int64
|
|
|
|
let
|
|
|
|
rdb = xDbs.cdb[6].rocksStoreRef
|
|
|
|
|
|
|
|
# Note that 32 and 33 size keys cannot be usefiully merged into the
|
|
|
|
# same SST file. The keys must be added in a sorted mode. So playing
|
|
|
|
# safe, key sizes should be of
|
|
|
|
# equal length.
|
|
|
|
|
|
|
|
if noisy: echo ""
|
|
|
|
noisy.showElapsed("Rocky bulk loader(keyLen 32)", ela[0]):
|
|
|
|
let bulker = RockyBulkLoadRef.init(rdb)
|
|
|
|
defer: bulker.destroy()
|
|
|
|
check bulker.begin("rocky-bulk-cache")
|
|
|
|
|
|
|
|
var
|
|
|
|
keyList = newSeq[NodeTag](xTab32.len)
|
|
|
|
|
|
|
|
fullNoise.showElapsed("Rocky bulk loader/32, sorter", ela[1]):
|
|
|
|
var inx = 0
|
|
|
|
for key in xTab32.keys:
|
|
|
|
keyList[inx] = key.to(NodeTag)
|
|
|
|
inx.inc
|
|
|
|
keyList.sort(cmp)
|
|
|
|
|
|
|
|
fullNoise.showElapsed("Rocky bulk loader/32, append", ela[2]):
|
|
|
|
for n,nodeTag in keyList:
|
|
|
|
let key = nodeTag.to(Blob)
|
|
|
|
check bulker.add(key, xTab32[key.to(ByteArray32)])
|
|
|
|
|
|
|
|
fullNoise.showElapsed("Rocky bulk loader/32, slurp", ela[3]):
|
|
|
|
let rc = bulker.finish()
|
|
|
|
if rc.isOk:
|
|
|
|
size = rc.value
|
|
|
|
else:
|
|
|
|
check bulker.lastError == "" # force printing error
|
|
|
|
|
|
|
|
fullNoise.say "***", " ela[]=", $ela.toSeq.mapIt(it.pp)
|
|
|
|
if ela[0].inNanoseconds != 0:
|
|
|
|
let
|
|
|
|
elaNs = ela.toSeq.mapIt(it.inNanoseconds.float)
|
|
|
|
elaPc = elaNs.mapIt(((it / elaNs[0]) * 100 + 0.5).int)
|
|
|
|
perRec = ((elaNs[0] / xTab32.len.float) + 0.5).int.initDuration
|
|
|
|
noisy.say "***",
|
|
|
|
"nRecords=", xTab32.len, ", ",
|
|
|
|
"perRecord=", perRec.pp, ", ",
|
|
|
|
"sstSize=", size.uint64.toSI, ", ",
|
|
|
|
"perRecord=", ((size.float / xTab32.len.float) + 0.5).int, ", ",
|
|
|
|
["Total","Sorter","Append","Ingest"].zip(elaPc).ppKvPc
|
|
|
|
|
|
|
|
# cdb[7] -- rocksdb, key length 32 as 33
|
|
|
|
test &"Store {xTab32.len} records " &
|
|
|
|
"(key length 33) into empty rocksdb table":
|
|
|
|
var
|
|
|
|
ela: array[4,Duration]
|
|
|
|
size: int64
|
|
|
|
let
|
|
|
|
rdb = xDbs.cdb[7].rocksStoreRef
|
|
|
|
|
|
|
|
# Note that 32 and 33 size keys cannot be usefiully merged into the
|
|
|
|
# same SST file. The keys must be added in a sorted mode. So playing
|
|
|
|
# safe, key sizes should be of
|
|
|
|
# equal length.
|
|
|
|
|
|
|
|
if noisy: echo ""
|
|
|
|
noisy.showElapsed("Rocky bulk loader(keyLen 32 as 33)", ela[0]):
|
|
|
|
let bulker = RockyBulkLoadRef.init(rdb)
|
|
|
|
defer: bulker.destroy()
|
|
|
|
check bulker.begin("rocky-bulk-cache")
|
|
|
|
|
|
|
|
var
|
|
|
|
keyList = newSeq[NodeTag](xTab32.len)
|
|
|
|
|
|
|
|
fullNoise.showElapsed("Rocky bulk loader/32 as 33, sorter", ela[1]):
|
|
|
|
var inx = 0
|
|
|
|
for key in xTab32.keys:
|
|
|
|
keyList[inx] = key.to(NodeTag)
|
|
|
|
inx.inc
|
|
|
|
keyList.sort(cmp)
|
|
|
|
|
|
|
|
fullNoise.showElapsed("Rocky bulk loader/32 as 33, append", ela[2]):
|
|
|
|
for n,nodeTag in keyList:
|
|
|
|
let key = nodeTag.to(Blob)
|
|
|
|
check bulker.add(@[99.byte] & key, xTab32[key.to(ByteArray32)])
|
|
|
|
|
|
|
|
fullNoise.showElapsed("Rocky bulk loader/32 as 33, slurp", ela[3]):
|
|
|
|
let rc = bulker.finish()
|
|
|
|
if rc.isOk:
|
|
|
|
size = rc.value
|
|
|
|
else:
|
|
|
|
check bulker.lastError == "" # force printing error
|
|
|
|
|
|
|
|
fullNoise.say "***", " ela[]=", $ela.toSeq.mapIt(it.pp)
|
|
|
|
if ela[0].inNanoseconds != 0:
|
|
|
|
let
|
|
|
|
elaNs = ela.toSeq.mapIt(it.inNanoseconds.float)
|
|
|
|
elaPc = elaNs.mapIt(((it / elaNs[0]) * 100 + 0.5).int)
|
|
|
|
perRec = ((elaNs[0] / xTab32.len.float) + 0.5).int.initDuration
|
|
|
|
noisy.say "***",
|
|
|
|
"nRecords=", xTab32.len, ", ",
|
|
|
|
"perRecord=", perRec.pp, ", ",
|
|
|
|
"sstSize=", size.uint64.toSI, ", ",
|
|
|
|
"perRecord=", ((size.float / xTab32.len.float) + 0.5).int, ", ",
|
|
|
|
["Total","Sorter","Append","Ingest"].zip(elaPc).ppKvPc
|
|
|
|
|
|
|
|
|
|
|
|
# cdb[8] -- rocksdb, key length 33
|
|
|
|
test &"Store {xTab33.len} records " &
|
|
|
|
&"(key length 33) into {emptyDb} rocksdb table":
|
|
|
|
var
|
|
|
|
ela: array[4,Duration]
|
|
|
|
size: int64
|
|
|
|
let rdb = xDbs.cdb[8].rocksStoreRef
|
|
|
|
|
|
|
|
# Note that 32 and 33 size keys cannot be usefiully merged into the
|
|
|
|
# same SST file. The keys must be added in a sorted mode. So playing
|
|
|
|
# safe, key sizes should be of equal length.
|
|
|
|
|
|
|
|
if noisy: echo ""
|
|
|
|
noisy.showElapsed("Rocky bulk loader(keyLen 33)", ela[0]):
|
|
|
|
let bulker = RockyBulkLoadRef.init(rdb)
|
|
|
|
defer: bulker.destroy()
|
|
|
|
check bulker.begin("rocky-bulk-cache")
|
|
|
|
|
|
|
|
var
|
|
|
|
kKeys: seq[byte] # need to cacscade
|
|
|
|
kTab: Table[byte,seq[NodeTag]]
|
|
|
|
|
|
|
|
fullNoise.showElapsed("Rocky bulk loader/33, sorter", ela[1]):
|
|
|
|
for key in xTab33.keys:
|
|
|
|
if kTab.hasKey(key[0]):
|
|
|
|
kTab[key[0]].add key.toOpenArray(1,32).to(NodeTag)
|
|
|
|
else:
|
|
|
|
kTab[key[0]] = @[key.toOpenArray(1,32).to(NodeTag)]
|
|
|
|
|
|
|
|
kKeys = toSeq(kTab.keys).sorted
|
|
|
|
for w in kKeys:
|
|
|
|
kTab[w].sort(cmp)
|
|
|
|
|
|
|
|
fullNoise.showElapsed("Rocky bulk loader/33, append", ela[2]):
|
|
|
|
for w in kKeys:
|
|
|
|
fullNoise.say "***", " prefix=", w, " entries=", kTab[w].len
|
|
|
|
for n,nodeTag in kTab[w]:
|
|
|
|
let key = (w,nodeTag).to(Blob)
|
|
|
|
check bulker.add(key, xTab33[key.to(ByteArray33)])
|
|
|
|
|
|
|
|
fullNoise.showElapsed("Rocky bulk loader/33, slurp", ela[3]):
|
|
|
|
let rc = bulker.finish()
|
|
|
|
if rc.isOk:
|
|
|
|
size = rc.value
|
|
|
|
else:
|
|
|
|
check bulker.lastError == "" # force printing error
|
|
|
|
|
|
|
|
fullNoise.say "***", " ela[]=", $ela.toSeq.mapIt(it.pp)
|
|
|
|
if ela[0].inNanoseconds != 0:
|
|
|
|
let
|
|
|
|
elaNs = ela.toSeq.mapIt(it.inNanoseconds.float)
|
|
|
|
elaPc = elaNs.mapIt(((it / elaNs[0]) * 100 + 0.5).int)
|
|
|
|
perRec = ((elaNs[0] / xTab33.len.float) + 0.5).int.initDuration
|
|
|
|
noisy.say "***",
|
|
|
|
"nRecords=", xTab33.len, ", ",
|
|
|
|
"perRecord=", perRec.pp, ", ",
|
|
|
|
"sstSize=", size.uint64.toSI, ", ",
|
|
|
|
"perRecord=", ((size.float / xTab33.len.float) + 0.5).int, ", ",
|
|
|
|
["Total","Cascaded-Sorter","Append","Ingest"].zip(elaPc).ppKvPc
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
# Main function(s)
|
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
|
|
|
|
proc syncSnapMain*(noisy = defined(debug)) =
|
2022-08-15 15:51:50 +00:00
|
|
|
noisy.accountsRunner(persistent=true)
|
2022-09-16 07:24:12 +00:00
|
|
|
#noisy.accountsRunner(persistent=false) # problems unless running stand-alone
|
2022-08-12 15:42:07 +00:00
|
|
|
noisy.importRunner() # small sample, just verify functionality
|
2022-09-16 07:24:12 +00:00
|
|
|
noisy.inspectionRunner()
|
2022-08-12 15:42:07 +00:00
|
|
|
noisy.storeRunner()
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
when isMainModule:
|
2022-08-04 08:04:30 +00:00
|
|
|
const
|
|
|
|
noisy = defined(debug) or true
|
2022-08-17 07:30:11 +00:00
|
|
|
|
2022-08-12 15:42:07 +00:00
|
|
|
#setTraceLevel()
|
2022-08-04 08:04:30 +00:00
|
|
|
setErrorLevel()
|
2022-07-01 11:42:17 +00:00
|
|
|
|
2022-08-17 07:30:11 +00:00
|
|
|
# The `accountsRunner()` tests a snap sync functionality for storing chain
|
|
|
|
# chain data directly rather than derive them by executing the EVM. Here,
|
|
|
|
# only accounts are considered.
|
|
|
|
#
|
|
|
|
# The `snap/1` protocol allows to fetch data for a certain account range. The
|
|
|
|
# following boundary conditions apply to the received data:
|
|
|
|
#
|
|
|
|
# * `State root`: All data are relaive to the same state root.
|
|
|
|
#
|
|
|
|
# * `Accounts`: There is an accounts interval sorted in strictly increasing
|
|
|
|
# order. The accounts are required consecutive, i.e. without holes in
|
|
|
|
# between although this cannot be verified immediately.
|
|
|
|
#
|
|
|
|
# * `Lower bound`: There is a start value which might be lower than the first
|
|
|
|
# account hash. There must be no other account between this start value and
|
|
|
|
# the first account (not verifyable yet.) For all practicat purposes, this
|
|
|
|
# value is mostly ignored but carried through.
|
|
|
|
#
|
|
|
|
# * `Proof`: There is a list of hexary nodes which allow to build a partial
|
2022-10-20 16:59:54 +00:00
|
|
|
# Patricia-Merkle trie starting at the state root with all the account
|
2022-08-17 07:30:11 +00:00
|
|
|
# leaves. There are enough nodes that show that there is no account before
|
|
|
|
# the least account (which is currently ignored.)
|
2022-09-02 18:16:09 +00:00
|
|
|
#
|
2022-08-17 07:30:11 +00:00
|
|
|
# There are test data samples on the sub-directory `test_sync_snap`. These
|
|
|
|
# are complete replies for some (admittedly smapp) test requests from a `kiln`
|
|
|
|
# session.
|
|
|
|
#
|
|
|
|
# The `accountsRunner()` does three tests:
|
|
|
|
#
|
|
|
|
# 1. Run the `importAccounts()` function which is the all-in-one production
|
|
|
|
# function processoing the data described above. The test applies it
|
|
|
|
# sequentially to about 20 data sets.
|
|
|
|
#
|
|
|
|
# 2. Test individual functional items which are hidden in test 1. while
|
2022-09-02 18:16:09 +00:00
|
|
|
# merging the sample data.
|
2022-08-17 07:30:11 +00:00
|
|
|
# * Load/accumulate `proofs` data from several samples
|
|
|
|
# * Load/accumulate accounts (needs some unique sorting)
|
|
|
|
# * Build/complete hexary trie for accounts
|
|
|
|
# * Save/bulk-store hexary trie on disk. If rocksdb is available, data
|
2022-10-20 16:59:54 +00:00
|
|
|
# are bulk stored via sst.
|
2022-08-17 07:30:11 +00:00
|
|
|
#
|
|
|
|
# 3. Traverse trie nodes stored earlier. The accounts from test 2 are
|
|
|
|
# re-visted using the account hash as access path.
|
|
|
|
#
|
|
|
|
|
2022-09-02 18:16:09 +00:00
|
|
|
# This one uses dumps from the external `nimbus-eth1-blob` repo
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
when true and false:
|
2022-09-02 18:16:09 +00:00
|
|
|
import ./test_sync_snap/snap_other_xx
|
|
|
|
noisy.showElapsed("accountsRunner()"):
|
|
|
|
for n,sam in snapOtherList:
|
2022-09-16 07:24:12 +00:00
|
|
|
false.accountsRunner(persistent=true, sam)
|
|
|
|
noisy.showElapsed("inspectRunner()"):
|
|
|
|
for n,sam in snapOtherHealingList:
|
|
|
|
false.inspectionRunner(persistent=true, cascaded=false, sam)
|
2022-09-02 18:16:09 +00:00
|
|
|
|
|
|
|
# This one usues dumps from the external `nimbus-eth1-blob` repo
|
Prep for full sync after snap make 4 (#1282)
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
2022-10-27 13:49:28 +00:00
|
|
|
when true and false:
|
2022-09-02 18:16:09 +00:00
|
|
|
import ./test_sync_snap/snap_storage_xx
|
|
|
|
let knownFailures = @[
|
|
|
|
("storages3__18__25_dump#11", @[( 233, RightBoundaryProofFailed)]),
|
|
|
|
("storages4__26__33_dump#11", @[(1193, RightBoundaryProofFailed)]),
|
|
|
|
("storages5__34__41_dump#10", @[( 508, RootNodeMismatch)]),
|
|
|
|
("storagesB__84__92_dump#6", @[( 325, RightBoundaryProofFailed)]),
|
|
|
|
("storagesD_102_109_dump#17", @[(1102, RightBoundaryProofFailed)]),
|
|
|
|
]
|
|
|
|
noisy.showElapsed("storageRunner()"):
|
|
|
|
for n,sam in snapStorageList:
|
2022-09-16 07:24:12 +00:00
|
|
|
false.storagesRunner(persistent=true, sam, knownFailures)
|
2022-09-02 18:16:09 +00:00
|
|
|
|
|
|
|
# This one uses readily available dumps
|
2022-08-24 13:44:18 +00:00
|
|
|
when true: # and false:
|
2022-09-16 07:24:12 +00:00
|
|
|
false.inspectionRunner()
|
|
|
|
for sam in snapTestList:
|
2022-09-02 18:16:09 +00:00
|
|
|
false.accountsRunner(persistent=true, sam)
|
2022-09-16 07:24:12 +00:00
|
|
|
for sam in snapTestStorageList:
|
2022-09-02 18:16:09 +00:00
|
|
|
false.accountsRunner(persistent=true, sam)
|
|
|
|
false.storagesRunner(persistent=true, sam)
|
|
|
|
|
|
|
|
# This one uses readily available dumps
|
|
|
|
when true and false:
|
2022-08-15 15:51:50 +00:00
|
|
|
# ---- database storage timings -------
|
2022-08-12 15:42:07 +00:00
|
|
|
|
2022-08-15 15:51:50 +00:00
|
|
|
noisy.showElapsed("importRunner()"):
|
|
|
|
noisy.importRunner(capture = bulkTest0)
|
2022-08-12 15:42:07 +00:00
|
|
|
|
2022-08-15 15:51:50 +00:00
|
|
|
noisy.showElapsed("storeRunner()"):
|
|
|
|
true.storeRunner(cleanUp = false)
|
|
|
|
true.storeRunner()
|
2022-07-01 11:42:17 +00:00
|
|
|
|
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
# End
|
|
|
|
# ------------------------------------------------------------------------------
|