2023-05-14 17:43:01 +00:00
|
|
|
# nimbus-eth1
|
2024-06-07 21:39:58 +00:00
|
|
|
# Copyright (c) 2023-2024 Status Research & Development GmbH
|
2023-05-14 17:43:01 +00:00
|
|
|
# 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.
|
|
|
|
|
|
|
|
## In-memory backend for Aristo DB
|
|
|
|
## ===============================
|
2023-06-20 13:26:25 +00:00
|
|
|
##
|
|
|
|
## The iterators provided here are currently available only by direct
|
|
|
|
## backend access
|
|
|
|
## ::
|
|
|
|
## import
|
|
|
|
## aristo/aristo_init,
|
|
|
|
## aristo/aristo_init/aristo_memory
|
|
|
|
##
|
2023-08-07 17:45:23 +00:00
|
|
|
## let rc = newAristoDbRef(BackendMemory)
|
2023-06-20 13:26:25 +00:00
|
|
|
## if rc.isOk:
|
|
|
|
## let be = rc.value.to(MemBackendRef)
|
|
|
|
## for (n, key, vtx) in be.walkVtx:
|
|
|
|
## ...
|
|
|
|
##
|
2023-05-14 17:43:01 +00:00
|
|
|
{.push raises: [].}
|
|
|
|
|
|
|
|
import
|
2023-08-21 14:58:30 +00:00
|
|
|
std/[algorithm, options, sequtils, tables],
|
2023-06-20 13:26:25 +00:00
|
|
|
eth/common,
|
2023-09-12 18:45:12 +00:00
|
|
|
results,
|
2023-06-20 13:26:25 +00:00
|
|
|
../aristo_constants,
|
2023-06-12 13:48:47 +00:00
|
|
|
../aristo_desc,
|
2023-08-25 22:53:59 +00:00
|
|
|
../aristo_desc/desc_backend,
|
2023-11-08 12:18:32 +00:00
|
|
|
../aristo_blobify,
|
2023-08-25 22:53:59 +00:00
|
|
|
./init_common
|
2023-05-14 17:43:01 +00:00
|
|
|
|
2024-04-16 20:39:11 +00:00
|
|
|
const
|
2024-07-04 13:46:52 +00:00
|
|
|
extraTraceMessages = false # or true
|
2024-04-16 20:39:11 +00:00
|
|
|
## Enabled additional logging noise
|
|
|
|
|
2023-05-14 17:43:01 +00:00
|
|
|
type
|
2024-03-14 22:17:43 +00:00
|
|
|
MemDbRef = ref object
|
|
|
|
## Database
|
2024-10-01 21:03:10 +00:00
|
|
|
sTab: Table[RootedVertexID,seq[byte]] ## Structural vertex table making up a trie
|
|
|
|
kMap: Table[RootedVertexID,HashKey] ## Merkle hash key mapping
|
|
|
|
tUvi: Option[VertexID] ## Top used vertex ID
|
|
|
|
lSst: Opt[SavedState] ## Last saved state
|
2023-06-09 11:17:37 +00:00
|
|
|
|
2024-03-14 22:17:43 +00:00
|
|
|
MemBackendRef* = ref object of TypedBackendRef
|
|
|
|
## Inheriting table so access can be extended for debugging purposes
|
2024-10-01 21:03:10 +00:00
|
|
|
mdb: MemDbRef ## Database
|
2024-03-14 22:17:43 +00:00
|
|
|
|
2023-06-20 13:26:25 +00:00
|
|
|
MemPutHdlRef = ref object of TypedPutHdlRef
|
2024-10-01 21:03:10 +00:00
|
|
|
sTab: Table[RootedVertexID,seq[byte]]
|
2024-07-04 13:46:52 +00:00
|
|
|
kMap: Table[RootedVertexID,HashKey]
|
2024-06-04 15:05:13 +00:00
|
|
|
tUvi: Option[VertexID]
|
2024-06-07 21:39:58 +00:00
|
|
|
lSst: Opt[SavedState]
|
2023-08-10 20:01:28 +00:00
|
|
|
|
2024-04-16 20:39:11 +00:00
|
|
|
when extraTraceMessages:
|
|
|
|
import chronicles
|
|
|
|
|
|
|
|
logScope:
|
|
|
|
topics = "aristo-backend"
|
|
|
|
|
2023-06-20 13:26:25 +00:00
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
# Private helpers
|
|
|
|
# ------------------------------------------------------------------------------
|
2023-06-09 11:17:37 +00:00
|
|
|
|
2023-06-20 13:26:25 +00:00
|
|
|
proc newSession(db: MemBackendRef): MemPutHdlRef =
|
|
|
|
new result
|
|
|
|
result.TypedPutHdlRef.beginSession db
|
|
|
|
|
|
|
|
proc getSession(hdl: PutHdlRef; db: MemBackendRef): MemPutHdlRef =
|
|
|
|
hdl.TypedPutHdlRef.verifySession db
|
|
|
|
hdl.MemPutHdlRef
|
|
|
|
|
|
|
|
proc endSession(hdl: PutHdlRef; db: MemBackendRef): MemPutHdlRef =
|
|
|
|
hdl.TypedPutHdlRef.finishSession db
|
|
|
|
hdl.MemPutHdlRef
|
2023-05-14 17:43:01 +00:00
|
|
|
|
|
|
|
# ------------------------------------------------------------------------------
|
2023-06-20 13:26:25 +00:00
|
|
|
# Private functions: interface
|
2023-05-14 17:43:01 +00:00
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
|
|
|
|
proc getVtxFn(db: MemBackendRef): GetVtxFn =
|
|
|
|
result =
|
2024-09-20 05:43:53 +00:00
|
|
|
proc(rvid: RootedVertexID, flags: set[GetVtxFlag]): Result[VertexRef,AristoError] =
|
2023-08-18 19:46:55 +00:00
|
|
|
# Fetch serialised data record
|
2024-07-04 13:46:52 +00:00
|
|
|
let data = db.mdb.sTab.getOrDefault(rvid, EmptyBlob)
|
2023-08-18 19:46:55 +00:00
|
|
|
if 0 < data.len:
|
2024-04-16 20:39:11 +00:00
|
|
|
let rc = data.deblobify(VertexRef)
|
|
|
|
when extraTraceMessages:
|
|
|
|
if rc.isErr:
|
2024-07-04 13:46:52 +00:00
|
|
|
trace logTxt "getVtxFn() failed", error=rc.error
|
2023-08-18 19:46:55 +00:00
|
|
|
return rc
|
2023-06-20 13:26:25 +00:00
|
|
|
err(GetVtxNotFound)
|
2023-05-14 17:43:01 +00:00
|
|
|
|
|
|
|
proc getKeyFn(db: MemBackendRef): GetKeyFn =
|
|
|
|
result =
|
2024-07-04 13:46:52 +00:00
|
|
|
proc(rvid: RootedVertexID): Result[HashKey,AristoError] =
|
|
|
|
let key = db.mdb.kMap.getOrVoid rvid
|
2023-06-22 11:13:24 +00:00
|
|
|
if key.isValid:
|
2023-05-30 21:21:15 +00:00
|
|
|
return ok key
|
2023-06-20 13:26:25 +00:00
|
|
|
err(GetKeyNotFound)
|
2023-05-14 17:43:01 +00:00
|
|
|
|
2024-06-04 15:05:13 +00:00
|
|
|
proc getTuvFn(db: MemBackendRef): GetTuvFn =
|
2023-06-09 11:17:37 +00:00
|
|
|
result =
|
2024-06-04 15:05:13 +00:00
|
|
|
proc(): Result[VertexID,AristoError]=
|
|
|
|
if db.mdb.tUvi.isSome:
|
|
|
|
return ok db.mdb.tUvi.unsafeGet
|
|
|
|
err(GetTuvNotFound)
|
2023-06-09 11:17:37 +00:00
|
|
|
|
2024-05-31 17:32:22 +00:00
|
|
|
proc getLstFn(db: MemBackendRef): GetLstFn =
|
|
|
|
result =
|
|
|
|
proc(): Result[SavedState,AristoError]=
|
|
|
|
if db.mdb.lSst.isSome:
|
|
|
|
return ok db.mdb.lSst.unsafeGet
|
|
|
|
err(GetLstNotFound)
|
|
|
|
|
2023-06-09 11:17:37 +00:00
|
|
|
# -------------
|
|
|
|
|
|
|
|
proc putBegFn(db: MemBackendRef): PutBegFn =
|
|
|
|
result =
|
2024-06-13 18:15:11 +00:00
|
|
|
proc(): Result[PutHdlRef,AristoError] =
|
|
|
|
ok db.newSession()
|
2023-06-09 11:17:37 +00:00
|
|
|
|
|
|
|
|
2023-05-14 17:43:01 +00:00
|
|
|
proc putVtxFn(db: MemBackendRef): PutVtxFn =
|
|
|
|
result =
|
2024-07-04 13:46:52 +00:00
|
|
|
proc(hdl: PutHdlRef; rvid: RootedVertexID; vtx: VertexRef) =
|
2023-06-20 13:26:25 +00:00
|
|
|
let hdl = hdl.getSession db
|
2023-06-30 22:22:33 +00:00
|
|
|
if hdl.error.isNil:
|
2024-06-25 11:39:53 +00:00
|
|
|
if vtx.isValid:
|
2024-09-02 14:03:10 +00:00
|
|
|
hdl.sTab[rvid] = vtx.blobify()
|
2024-06-25 11:39:53 +00:00
|
|
|
else:
|
2024-07-04 13:46:52 +00:00
|
|
|
hdl.sTab[rvid] = EmptyBlob
|
2023-05-14 17:43:01 +00:00
|
|
|
|
|
|
|
proc putKeyFn(db: MemBackendRef): PutKeyFn =
|
|
|
|
result =
|
2024-07-04 13:46:52 +00:00
|
|
|
proc(hdl: PutHdlRef; rvid: RootedVertexID, key: HashKey) =
|
2023-06-20 13:26:25 +00:00
|
|
|
let hdl = hdl.getSession db
|
2023-06-30 22:22:33 +00:00
|
|
|
if hdl.error.isNil:
|
2024-07-04 13:46:52 +00:00
|
|
|
hdl.kMap[rvid] = key
|
2023-05-14 17:43:01 +00:00
|
|
|
|
2024-06-04 15:05:13 +00:00
|
|
|
proc putTuvFn(db: MemBackendRef): PutTuvFn =
|
2023-06-09 11:17:37 +00:00
|
|
|
result =
|
2024-06-04 15:05:13 +00:00
|
|
|
proc(hdl: PutHdlRef; vs: VertexID) =
|
2023-06-20 13:26:25 +00:00
|
|
|
let hdl = hdl.getSession db
|
2023-06-30 22:22:33 +00:00
|
|
|
if hdl.error.isNil:
|
2024-06-04 15:05:13 +00:00
|
|
|
hdl.tUvi = some(vs)
|
2023-06-09 11:17:37 +00:00
|
|
|
|
2024-05-31 17:32:22 +00:00
|
|
|
proc putLstFn(db: MemBackendRef): PutLstFn =
|
|
|
|
result =
|
|
|
|
proc(hdl: PutHdlRef; lst: SavedState) =
|
|
|
|
let hdl = hdl.getSession db
|
|
|
|
if hdl.error.isNil:
|
2024-06-05 18:17:50 +00:00
|
|
|
let rc = lst.blobify # test
|
|
|
|
if rc.isOk:
|
2024-06-07 21:39:58 +00:00
|
|
|
hdl.lSst = Opt.some(lst)
|
2024-06-05 18:17:50 +00:00
|
|
|
else:
|
|
|
|
hdl.error = TypedPutHdlErrRef(
|
|
|
|
pfx: AdmPfx,
|
|
|
|
aid: AdmTabIdLst,
|
|
|
|
code: rc.error)
|
2024-05-31 17:32:22 +00:00
|
|
|
|
2023-06-09 11:17:37 +00:00
|
|
|
proc putEndFn(db: MemBackendRef): PutEndFn =
|
|
|
|
result =
|
2023-09-12 18:45:12 +00:00
|
|
|
proc(hdl: PutHdlRef): Result[void,AristoError] =
|
2023-06-20 13:26:25 +00:00
|
|
|
let hdl = hdl.endSession db
|
2023-06-30 22:22:33 +00:00
|
|
|
if not hdl.error.isNil:
|
2024-04-16 20:39:11 +00:00
|
|
|
when extraTraceMessages:
|
|
|
|
case hdl.error.pfx:
|
|
|
|
of VtxPfx, KeyPfx: trace logTxt "putEndFn: vtx/key failed",
|
2023-08-18 19:46:55 +00:00
|
|
|
pfx=hdl.error.pfx, vid=hdl.error.vid, error=hdl.error.code
|
2024-04-16 20:39:11 +00:00
|
|
|
of AdmPfx: trace logTxt "putEndFn: admin failed",
|
|
|
|
pfx=AdmPfx, aid=hdl.error.aid.uint64, error=hdl.error.code
|
|
|
|
of Oops: trace logTxt "putEndFn: failed",
|
|
|
|
pfx=hdl.error.pfx, error=hdl.error.code
|
2023-09-12 18:45:12 +00:00
|
|
|
return err(hdl.error.code)
|
2023-06-20 13:26:25 +00:00
|
|
|
|
2023-08-18 19:46:55 +00:00
|
|
|
for (vid,data) in hdl.sTab.pairs:
|
|
|
|
if 0 < data.len:
|
2024-03-14 22:17:43 +00:00
|
|
|
db.mdb.sTab[vid] = data
|
2023-06-20 13:26:25 +00:00
|
|
|
else:
|
2024-03-14 22:17:43 +00:00
|
|
|
db.mdb.sTab.del vid
|
2023-06-20 13:26:25 +00:00
|
|
|
|
|
|
|
for (vid,key) in hdl.kMap.pairs:
|
|
|
|
if key.isValid:
|
2024-03-14 22:17:43 +00:00
|
|
|
db.mdb.kMap[vid] = key
|
2023-06-20 13:26:25 +00:00
|
|
|
else:
|
2024-03-14 22:17:43 +00:00
|
|
|
db.mdb.kMap.del vid
|
2023-06-20 13:26:25 +00:00
|
|
|
|
2024-06-04 15:05:13 +00:00
|
|
|
let tuv = hdl.tUvi.get(otherwise = VertexID(0))
|
|
|
|
if tuv.isValid:
|
|
|
|
db.mdb.tUvi = some(tuv)
|
2023-08-22 18:44:54 +00:00
|
|
|
|
2024-05-31 17:32:22 +00:00
|
|
|
if hdl.lSst.isSome:
|
|
|
|
db.mdb.lSst = hdl.lSst
|
|
|
|
|
2023-09-12 18:45:12 +00:00
|
|
|
ok()
|
2023-06-09 11:17:37 +00:00
|
|
|
|
|
|
|
# -------------
|
|
|
|
|
2023-06-20 13:26:25 +00:00
|
|
|
proc closeFn(db: MemBackendRef): CloseFn =
|
2023-05-14 17:43:01 +00:00
|
|
|
result =
|
2023-06-20 13:26:25 +00:00
|
|
|
proc(ignore: bool) =
|
|
|
|
discard
|
2023-05-14 17:43:01 +00:00
|
|
|
|
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
# Public functions
|
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
|
2024-06-03 20:10:35 +00:00
|
|
|
proc memoryBackend*(): BackendRef =
|
2023-09-05 13:57:20 +00:00
|
|
|
let db = MemBackendRef(
|
|
|
|
beKind: BackendMemory,
|
2024-03-14 22:17:43 +00:00
|
|
|
mdb: MemDbRef())
|
|
|
|
|
2023-06-20 13:26:25 +00:00
|
|
|
db.getVtxFn = getVtxFn db
|
|
|
|
db.getKeyFn = getKeyFn db
|
2024-06-04 15:05:13 +00:00
|
|
|
db.getTuvFn = getTuvFn db
|
2024-05-31 17:32:22 +00:00
|
|
|
db.getLstFn = getLstFn db
|
2023-06-20 13:26:25 +00:00
|
|
|
|
|
|
|
db.putBegFn = putBegFn db
|
|
|
|
db.putVtxFn = putVtxFn db
|
|
|
|
db.putKeyFn = putKeyFn db
|
2024-06-04 15:05:13 +00:00
|
|
|
db.putTuvFn = putTuvFn db
|
2024-05-31 17:32:22 +00:00
|
|
|
db.putLstFn = putLstFn db
|
2023-06-20 13:26:25 +00:00
|
|
|
db.putEndFn = putEndFn db
|
2023-05-14 17:43:01 +00:00
|
|
|
|
2023-06-20 13:26:25 +00:00
|
|
|
db.closeFn = closeFn db
|
|
|
|
db
|
|
|
|
|
2024-03-14 22:17:43 +00:00
|
|
|
proc dup*(db: MemBackendRef): MemBackendRef =
|
|
|
|
## Duplicate descriptor shell as needed for API debugging
|
|
|
|
new result
|
|
|
|
init_common.init(result[], db[])
|
|
|
|
result.mdb = db.mdb
|
|
|
|
|
2023-06-20 13:26:25 +00:00
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
# Public iterators (needs direct backend access)
|
|
|
|
# ------------------------------------------------------------------------------
|
2023-06-09 11:17:37 +00:00
|
|
|
|
2023-06-20 13:26:25 +00:00
|
|
|
iterator walkVtx*(
|
|
|
|
be: MemBackendRef;
|
Speed up initial MPT root computation after import (#2788)
When `nimbus import` runs, we end up with a database without MPT roots
leading to long startup times the first time one is needed.
Computing the state root is slow because the on-disk order based on
VertexID sorting does not match the trie traversal order and therefore
makes lookups inefficent.
Here we introduce a helper that speeds up this computation by traversing
the trie in on-disk order and computing the trie hashes bottom up
instead - even though this leads to some redundant reads of nodes that
we cannot yet compute, it's still a net win as leaves and "bottom"
branches make up the majority of the database.
This PR also addresses a few other sources of inefficiency largely due
to the separation of AriKey and AriVtx into their own column families.
Each column family is its own LSM tree that produces hundreds of SST
filtes - with a limit of 512 open files, rocksdb must keep closing and
opening files which leads to expensive metadata reads during random
access.
When rocksdb makes a lookup, it has to read several layers of files for
each lookup. Ribbon filters to skip over files that don't have the
requested data but when these filters are not in memory, reading them is
slow - this happens in two cases: when opening a file and when the
filter has been evicted from the LRU cache. Addressing the open file
limit solves one source of inefficiency, but we must also increase the
block cache size to deal with this problem.
* rocksdb.max_open_files increased to 2048
* per-file size limits increased so that fewer files are created
* WAL size increased to avoid partial flushes which lead to small files
* rocksdb block cache increased
All these increases of course lead to increased memory usage, but at
least performance is acceptable - in the future, we'll need to explore
options such as joining AriVtx and AriKey and/or reducing the row count
(by grouping branch layers under a single vertexid).
With this PR, the mainnet state root can be computed in ~8 hours (down
from 2-3 days) - not great, but still better.
Further, we write all keys to the database, also those that are less
than 32 bytes - because the mpt path is part of the input, it is very
rare that we actually hit a key like this (about 200k such entries on
mainnet), so the code complexity is not worth the benefit really, in the
current database layout / design.
2024-10-27 11:08:37 +00:00
|
|
|
kinds = {Branch, Leaf};
|
2024-07-04 13:46:52 +00:00
|
|
|
): tuple[rvid: RootedVertexID, vtx: VertexRef] =
|
2023-06-20 13:26:25 +00:00
|
|
|
## Iteration over the vertex sub-table.
|
2024-07-04 13:46:52 +00:00
|
|
|
for n,rvid in be.mdb.sTab.keys.toSeq.mapIt(it).sorted:
|
|
|
|
let data = be.mdb.sTab.getOrDefault(rvid, EmptyBlob)
|
2023-08-18 19:46:55 +00:00
|
|
|
if 0 < data.len:
|
|
|
|
let rc = data.deblobify VertexRef
|
|
|
|
if rc.isErr:
|
2024-04-16 20:39:11 +00:00
|
|
|
when extraTraceMessages:
|
2024-07-04 13:46:52 +00:00
|
|
|
debug logTxt "walkVtxFn() skip", n, rvid, error=rc.error
|
2023-08-18 19:46:55 +00:00
|
|
|
else:
|
Speed up initial MPT root computation after import (#2788)
When `nimbus import` runs, we end up with a database without MPT roots
leading to long startup times the first time one is needed.
Computing the state root is slow because the on-disk order based on
VertexID sorting does not match the trie traversal order and therefore
makes lookups inefficent.
Here we introduce a helper that speeds up this computation by traversing
the trie in on-disk order and computing the trie hashes bottom up
instead - even though this leads to some redundant reads of nodes that
we cannot yet compute, it's still a net win as leaves and "bottom"
branches make up the majority of the database.
This PR also addresses a few other sources of inefficiency largely due
to the separation of AriKey and AriVtx into their own column families.
Each column family is its own LSM tree that produces hundreds of SST
filtes - with a limit of 512 open files, rocksdb must keep closing and
opening files which leads to expensive metadata reads during random
access.
When rocksdb makes a lookup, it has to read several layers of files for
each lookup. Ribbon filters to skip over files that don't have the
requested data but when these filters are not in memory, reading them is
slow - this happens in two cases: when opening a file and when the
filter has been evicted from the LRU cache. Addressing the open file
limit solves one source of inefficiency, but we must also increase the
block cache size to deal with this problem.
* rocksdb.max_open_files increased to 2048
* per-file size limits increased so that fewer files are created
* WAL size increased to avoid partial flushes which lead to small files
* rocksdb block cache increased
All these increases of course lead to increased memory usage, but at
least performance is acceptable - in the future, we'll need to explore
options such as joining AriVtx and AriKey and/or reducing the row count
(by grouping branch layers under a single vertexid).
With this PR, the mainnet state root can be computed in ~8 hours (down
from 2-3 days) - not great, but still better.
Further, we write all keys to the database, also those that are less
than 32 bytes - because the mpt path is part of the input, it is very
rare that we actually hit a key like this (about 200k such entries on
mainnet), so the code complexity is not worth the benefit really, in the
current database layout / design.
2024-10-27 11:08:37 +00:00
|
|
|
if rc.value.vType in kinds:
|
|
|
|
yield (rvid, rc.value)
|
2023-06-20 13:26:25 +00:00
|
|
|
|
|
|
|
iterator walkKey*(
|
|
|
|
be: MemBackendRef;
|
2024-07-04 13:46:52 +00:00
|
|
|
): tuple[rvid: RootedVertexID, key: HashKey] =
|
2023-06-20 13:26:25 +00:00
|
|
|
## Iteration over the Markle hash sub-table.
|
2024-07-04 13:46:52 +00:00
|
|
|
for rvid in be.mdb.kMap.keys.toSeq.mapIt(it).sorted:
|
|
|
|
let key = be.mdb.kMap.getOrVoid(rvid)
|
2023-06-20 13:26:25 +00:00
|
|
|
if key.isValid:
|
2024-07-04 13:46:52 +00:00
|
|
|
yield (rvid, key)
|
2023-06-20 13:26:25 +00:00
|
|
|
|
2023-08-22 18:44:54 +00:00
|
|
|
|
2023-05-14 17:43:01 +00:00
|
|
|
# ------------------------------------------------------------------------------
|
|
|
|
# End
|
|
|
|
# ------------------------------------------------------------------------------
|