ARCHIVE/xahaud
3
0
Fork 0
mirror of https://github.com/Xahau/xahaud.git synced 2025-11-04 10:45:50 +00:00
Code Issues Packages Projects Releases Wiki Activity

Improve online_delete configuration and DB tuning:

Browse Source 
* Document delete_batch, back_off_milliseconds, age_threshold_seconds.
* Convert those time values to chrono types.
* Fix bug that ignored age_threshold_seconds.
* Add a "recovery buffer" to the config that gives the node a chance to
  recover before aborting online delete.
* Add begin/end log messages around the SQL queries.
* Add a new configuration section: [sqlite] to allow tuning the sqlite
  database operations. Ignored on full/large history servers.
* Update documentation of [node_db] and [sqlite] in the
  rippled-example.cfg file.

Resolves #3321
This commit is contained in:
Edward Hennis
2020-05-11 16:48:34 -04:00
committed by Nik Bougalis
parent 00702f28c2
commit 4702c8b591
21 changed files with 1086 additions and 271 deletions
Download Patch File Download Diff File Expand all files Collapse all files

181
cfg/rippled-example.cfg
View File

@@ -36,7 +36,7 @@
# For more information on where the rippled server instance searches for the
# file, visit:
#
# https://developers.ripple.com/commandline-usage.html#generic-options
# https://xrpl.org/commandline-usage.html#generic-options
#
# This file should be named rippled.cfg. This file is UTF-8 with DOS, UNIX,
# or Mac style end of lines. Blank lines and lines beginning with '#' are
@@ -869,18 +869,65 @@
#
# These keys are possible for any type of backend:
#
# earliest_seq The default is 32570 to match the XRP ledger
# network's earliest allowed sequence. Alternate
# networks may set this value. Minimum value of 1.
# If a [shard_db] section is defined, and this
# value is present either [node_db] or [shard_db],
# it must be defined with the same value in both
# sections.
#
# online_delete Minimum value of 256. Enable automatic purging
# of older ledger information. Maintain at least this
# number of ledger records online. Must be greater
# than or equal to ledger_history.
#
# advisory_delete 0 for disabled, 1 for enabled. If set, then
# require administrative RPC call "can_delete"
# to enable online deletion of ledger records.
# These keys modify the behavior of online_delete, and thus are only
# relevant if online_delete is defined and non-zero:
#
# earliest_seq The default is 32570 to match the XRP ledger
# network's earliest allowed sequence. Alternate
# networks may set this value. Minimum value of 1.
# advisory_delete 0 for disabled, 1 for enabled. If set, the
# administrative RPC call "can_delete" is required
# to enable online deletion of ledger records.
# Online deletion does not run automatically if
# non-zero and the last deletion was on a ledger
# greater than the current "can_delete" setting.
# Default is 0.
#
# delete_batch When automatically purging, SQLite database
# records are deleted in batches. This value
# controls the maximum size of each batch. Larger
# batches keep the databases locked for more time,
# which may cause other functions to fall behind,
# and thus cause the node to lose sync.
# Default is 100.
#
# back_off_milliseconds
# Number of milliseconds to wait between
# online_delete batches to allow other functions
# to catch up.
# Default is 100.
#
# age_threshold_seconds
# The online delete process will only run if the
# latest validated ledger is younger than this
# number of seconds.
# Default is 60.
#
# recovery_wait_seconds
# The online delete process checks periodically
# that rippled is still in sync with the network,
# and that the validated ledger is less than
# 'age_threshold_seconds' old. By default, if it
# is not the online delete process aborts and
# tries again later. If 'recovery_wait_seconds'
# is set and rippled is out of sync, but likely to
# recover quickly, then online delete will wait
# this number of seconds for rippled to get back
# into sync before it aborts.
# Set this value if the node is otherwise staying
# in sync, or recovering quickly, but the online
# delete process is unable to finish.
# Default is unset.
#
# Notes:
# The 'node_db' entry configures the primary, persistent storage.
@@ -892,6 +939,12 @@
# [import_db] Settings for performing a one-time import (optional)
# [database_path] Path to the book-keeping databases.
#
# The server creates and maintains 4 to 5 bookkeeping SQLite databases in
# the 'database_path' location. If you omit this configuration setting,
# the server creates a directory called "db" located in the same place as
# your rippled.cfg file.
# Partial pathnames are relative to the location of the rippled executable.
#
# [shard_db] Settings for the Shard Database (optional)
#
# Format (without spaces):
@@ -907,12 +960,84 @@
#
# max_size_gb Maximum disk space the database will utilize (in gigabytes)
#
# [sqlite] Tuning settings for the SQLite databases (optional)
#
# There are 4 bookkeeping SQLite database that the server creates and
# maintains. If you omit this configuration setting, it will default to
# creating a directory called "db" located in the same place as your
# rippled.cfg file. Partial pathnames will be considered relative to
# the location of the rippled executable.
# Format (without spaces):
# One or more lines of case-insensitive key / value pairs:
# <key> '=' <value>
# ...
#
# Example 1:
# sync_level=low
#
# Example 2:
# journal_mode=off
# synchronous=off
#
# WARNING: These settings can have significant effects on data integrity,
# particularly in systemic failure scenarios. It is strongly recommended
# that they be left at their defaults unless the server is having
# performance issues during normal operation or during automatic purging
# (online_delete) operations. A warning will be logged on startup if
# 'ledger_history' is configured to store more than 10,000,000 ledgers and
# any of these settings are less safe than the default. This is due to the
# inordinate amount of time and bandwidth it will take to safely rebuild a
# corrupted database of that size from other peers.
#
# Optional keys:
#
# safety_level Valid values: high, low
# The default is "high", which tunes the SQLite
# databases in the most reliable mode, and is
# equivalent to:
# journal_mode=wal
# synchronous=normal
# temp_store=file
# "low" is equivalent to:
# journal_mode=memory
# synchronous=off
# temp_store=memory
# These "low" settings trade speed and reduced I/O
# for a higher risk of data loss. See the
# individual settings below for more information.
# This setting may not be combined with any of the
# other tuning settings: "journal_mode",
# "synchronous", or "temp_store".
#
# journal_mode Valid values: delete, truncate, persist, memory, wal, off
# The default is "wal", which uses a write-ahead
# log to implement database transactions.
# Alternately, "memory" saves disk I/O, but if
# rippled crashes during a transaction, the
# database is likely to be corrupted.
# See https://www.sqlite.org/pragma.html#pragma_journal_mode
# for more details about the available options.
# This setting may not be combined with the
# "safety_level" setting.
#
# synchronous Valid values: off, normal, full, extra
# The default is "normal", which works well with
# the "wal" journal mode. Alternatively, "off"
# allows rippled to continue as soon as data is
# passed to the OS, which can significantly
# increase speed, but risks data corruption if
# the host computer crashes before writing that
# data to disk.
# See https://www.sqlite.org/pragma.html#pragma_synchronous
# for more details about the available options.
# This setting may not be combined with the
# "safety_level" setting.
#
# temp_store Valid values: default, file, memory
# The default is "file", which will use files
# for temporary database tables and indices.
# Alternatively, "memory" may save I/O, but
# rippled does not currently use many, if any,
# of these temporary objects.
# See https://www.sqlite.org/pragma.html#pragma_temp_store
# for more details about the available options.
# This setting may not be combined with the
# "safety_level" setting.
#
#
#
@@ -1212,24 +1337,27 @@ medium
# This is primary persistent datastore for rippled. This includes transaction
# metadata, account states, and ledger headers. Helpful information can be
# found here: https://ripple.com/wiki/NodeBackEnd
# delete old ledgers while maintaining at least 2000. Do not require an
# external administrative command to initiate deletion.
# found at https://xrpl.org/capacity-planning.html#node-db-type
# type=NuDB is recommended for non-validators with fast SSDs. Validators or
# slow / spinning disks should use RocksDB. Caution: Spinning disks are
# not recommended. They do not perform well enough to consistently remain
# synced to the network.
# online_delete=512 is recommended to delete old ledgers while maintaining at
# least 512.
# advisory_delete=0 allows the online delete process to run automatically
# when the node has approximately two times the "online_delete" value of
# ledgers. No external administrative command is required to initiate
# deletion.
[node_db]
type=RocksDB
path=/var/lib/rippled/db/rocksdb
open_files=2000
filter_bits=12
cache_mb=256
file_size_mb=8
file_size_mult=2
online_delete=2000
type=NuDB
path=/var/lib/rippled/db/nudb
online_delete=512
advisory_delete=0
# This is the persistent datastore for shards. It is important for the health
# of the ripple network that rippled operators shard as much as practical.
# NuDB requires SSD storage. Helpful information can be found here
# https://ripple.com/build/history-sharding
# NuDB requires SSD storage. Helpful information can be found at
# https://xrpl.org/history-sharding.html
#[shard_db]
#path=/var/lib/rippled/db/shards/nudb
#max_size_gb=500
@@ -1248,7 +1376,8 @@ time.apple.com
time.nist.gov
pool.ntp.org
# To use the XRP test network (see https://ripple.com/build/xrp-test-net/),
# To use the XRP test network
# (see https://xrpl.org/connect-your-rippled-to-the-xrp-test-net.html),
# use the following [ips] section:
# [ips]
# r.altnet.rippletest.net 51235