MaxWelll之生产配置案列以及注意事项

1 生产使用以及注意事项

生产MaxWell的版本是1.24.1,使用时有如下注意事项。

  • 时区相差8小时:TIMESTAMP类型数据采集时,会有八小时的时差,这个只能通过更改源码的方式解决
  • Bootstrap命令行BUG:当多数据库实例的采集配置公用同一MaxWell配置库时,使用Bootstrap命令行导入全量数据会有BUG,应使用SQL插入方式
  • server_id冲突BUG:maxwell是根据采集的mysql实例库的server_id来区分不同的实例的配置的,当采集的多个实例的server_id存在相同时则会使binglog读取串了
  • Bootstrap时区设置BUG:由于我们在东八区,故需要在jdbc_options添加时区信息,注意时区必须设置为serverTimezone=GMT+8,不设置以及设置为serverTimezone=Asia/Shanghai 都会有错误
  • kafka生产者尝试次数:默认kafka的尝试次数是0,当出现发送失败时就会使maxwell进程停止,需调整,设置为100以上,增加容错概率。
  • 不支持模式匹配分库分表:不支持类似于cannal那样模式匹配分库分表,如database*.table* 无法匹配到database1.table1、database2.table2,以及kafka输出时将分库分表合在一个topic中,若想实现此功能需要二次开发源码。

2 生产配置案列

如下是生产上maxwell的配置案列,实时采集指定的库以及表发送到特定topic中。

# tl;dr config
log_level=info

client_id=192.168.17.84

producer=kafka
kafka.bootstrap.servers=essrvs-kafka01.dc.servyou-it.com:9092,essrvs-kafka02.dc.servyou-it.com:9092,essrvs-kafka03.dc.servyou-it.com:9092

# mysql login info
host=192.168.17.94
user=maxwell
password=******
schema_database=maxwell


#     *** general ***
# choose where to produce data to. stdout|file|kafka|kinesis|pubsub|sqs|rabbitmq|redis
#producer=kafka

# set the log level.  note that you can configure things further in log4j2.xml
#log_level=DEBUG # [DEBUG, INFO, WARN, ERROR]

# if set, maxwell will look up the scoped environment variables, strip off the prefix and inject the configs
#env_config_prefix=MAXWELL_

#     *** mysql ***

# mysql host to connect to
#host=hostname

# mysql port to connect to
#port=3306

# mysql user to connect as.  This user must have REPLICATION SLAVE permissions,
# as well as full access to the `maxwell` (or schema_database) database
#user=maxwell

# mysql password
#password=maxwell

# options to pass into the jdbc connection, given as opt=val&opt2=val2
jdbc_options=characterEncoding=utf8&autoReconnect=true&allowMultiQueries=true&useSSL=false&serverTimezone=GMT+8
replication_jdbc_options=characterEncoding=utf8&autoReconnect=true&allowMultiQueries=true&useSSL=false&serverTimezone=GMT+8
schema_jdbc_options=characterEncoding=utf8&autoReconnect=true&allowMultiQueries=true&useSSL=false&serverTimezone=GMT+8

# name of the mysql database where maxwell keeps its own state
#schema_database=maxwell

# whether to use GTID or not for positioning
#gtid_mode=true

# SSL/TLS options
# To use VERIFY_CA or VERIFY_IDENTITY, you must set the trust store with Java opts:
#   -Djavax.net.ssl.trustStore= -Djavax.net.ssl.trustStorePassword=
# or import the MySQL cert into the global Java cacerts.
# MODE must be one of DISABLED, PREFERRED, REQUIRED, VERIFY_CA, or VERIFY_IDENTITY
#
# turns on ssl for the maxwell-store connection, other connections inherit this setting unless specified
#ssl=DISABLED
# for binlog-connector
#replication_ssl=DISABLED
# for the schema-capture connection, if used
#schema_ssl=DISABLED

# maxwell can optionally replicate from a different server than where it stores
# schema and binlog position info.  Specify that different server here:

replication_host=192.168.17.84
replication_user=rsync_user
replication_password=*******
replication_port=3377

# This may be useful when using MaxScale's binlog mirroring host.
# Specifies that Maxwell should capture schema from a different server than
# it replicates from:

#schema_host=other
#schema_user=username
#schema_password=password
#schema_port=3306

#       *** output format ***

# records include binlog position (default false)
#output_binlog_position=true

# records include a gtid string (default false)
#output_gtid_position=true

# records include fields with null values (default true).  If this is false,
# fields where the value is null will be omitted entirely from output.
#output_nulls=true

# records include server_id (default false)
#output_server_id=true

# records include thread_id (default false)
#output_thread_id=true

# records include schema_id (default false)
#output_schema_id=true

# records include row query, binlog option "binlog_rows_query_log_events" must be enabled" (default false)
#output_row_query=true

# DML records include list of values that make up a row's primary key (default false)
#output_primary_keys=true

# DML records include list of columns that make up a row's primary key (default false)
#output_primary_key_columns=true

# records include commit and xid (default true)
#output_commit_info=true

# This controls whether maxwell will output JSON information containing
# DDL (ALTER/CREATE TABLE/ETC) infromation. (default: false)
# See also: ddl_kafka_topic
#output_ddl=true

#       *** kafka ***

# list of kafka brokers
#kafka.bootstrap.servers=hosta:9092,hostb:9092

# kafka topic to write to
# this can be static, e.g. 'maxwell', or dynamic, e.g. namespace_%{database}_%{table}
# in the latter case 'database' and 'table' will be replaced with the values for the row being processed
kafka_topic=bdps_maxwell_%{database}_%{table}

# alternative kafka topic to write DDL (alter/create/drop) to.  Defaults to kafka_topic
#ddl_kafka_topic=maxwell_ddl

# hash function to use.  "default" is just the JVM's 'hashCode' function.
#kafka_partition_hash=default # [default, murmur3]

# how maxwell writes its kafka key.
#
# 'hash' looks like:
# {"database":"test","table":"tickets","pk.id":10001}
#
# 'array' looks like:
# ["test","tickets",[{"id":10001}]]
#
# default: "hash"
#kafka_key_format=hash # [hash, array]

# extra kafka options.  Anything prefixed "kafka." will get
# passed directly into the kafka-producer's config.

# a few defaults.
# These are 0.11-specific. They may or may not work with other versions.
# kafka.compression.type=snappy
kafka.retries=2000
kafka.acks=all
kafka.max.in.flight.requests.per.connection=1
#kafka.batch.size=16384


# kafka+SSL example
# kafka.security.protocol=SSL
# kafka.ssl.truststore.location=/var/private/ssl/kafka.client.truststore.jks
# kafka.ssl.truststore.password=test1234
# kafka.ssl.keystore.location=/var/private/ssl/kafka.client.keystore.jks
# kafka.ssl.keystore.password=test1234
# kafka.ssl.key.password=test1234#

# controls a heuristic check that maxwell may use to detect messages that
# we never heard back from.  The heuristic check looks for "stuck" messages, and
# will timeout maxwell after this many milliseconds.
#
# See https://github.com/zendesk/maxwell/blob/master/src/main/java/com/zendesk/maxwell/producer/InflightMessageList.java
# if you really want to get into it.
#producer_ack_timeout=120000 # default 0
ignore_producer_error=false


#           *** partitioning ***

# What part of the data do we partition by?
producer_partition_by=primary_key

# specify what fields to partition by when using producer_partition_by=column
# column separated list.
#producer_partition_columns=id,foo,bar

# when using producer_partition_by=column, partition by this when
# the specified column(s) don't exist.
#producer_partition_by_fallback=database

#            *** kinesis ***

#kinesis_stream=maxwell

# AWS places a 256 unicode character limit on the max key length of a record
# http://docs.aws.amazon.com/kinesis/latest/APIReference/API_PutRecord.html
#
# Setting this option to true enables hashing the key with the md5 algorithm
# before we send it to kinesis so all the keys work within the key size limit.
# Values: true, false
# Default: false
#kinesis_md5_keys=true

#            *** sqs ***

#sqs_queue_uri=aws_sqs_queue_uri

# The sqs producer will need aws credentials configured in the default
# root folder and file format. Please check below link on how to do it.
# http://docs.aws.amazon.com/sdk-for-java/v1/developer-guide/setup-credentials.html

#            *** pub/sub ***

#pubsub_project_id=maxwell
#pubsub_topic=maxwell
#ddl_pubsub_topic=maxwell_ddl

#            *** rabbit-mq ***

#rabbitmq_host=rabbitmq_hostname
#rabbitmq_port=5672
#rabbitmq_user=guest
#rabbitmq_pass=guest
#rabbitmq_virtual_host=/
#rabbitmq_exchange=maxwell
#rabbitmq_exchange_type=fanout
#rabbitmq_exchange_durable=false
#rabbitmq_exchange_autodelete=false
#rabbitmq_routing_key_template=%db%.%table%
#rabbitmq_message_persistent=false
#rabbitmq_declare_exchange=true

#           *** redis ***

#redis_host=redis_host
#redis_port=6379
#redis_auth=redis_auth
#redis_database=0

# name of pubsub/list/whatever key to publish to
#redis_key=maxwell

# this can be static, e.g. 'maxwell', or dynamic, e.g. namespace_%{database}_%{table}
#redis_pub_channel=maxwell
# this can be static, e.g. 'maxwell', or dynamic, e.g. namespace_%{database}_%{table}
#redis_list_key=maxwell
# this can be static, e.g. 'maxwell', or dynamic, e.g. namespace_%{database}_%{table}
# Valid values for redis_type = pubsub|lpush. Defaults to pubsub

#redis_type=pubsub

#           *** custom producer ***

# the fully qualified class name for custom ProducerFactory
# see the following link for more details.
# http://maxwells-daemon.io/producers/#custom-producer
#custom_producer.factory=

# custom producer properties can be configured using the custom_producer.* property namespace
#custom_producer.custom_prop=foo

#          *** filtering ***

# filter rows out of Maxwell's output.  Command separated list of filter-rules, evaluated in sequence.
# A filter rule is:
#   ":"  "."  [ "."  "="  ]
#  type    ::= [ "include" | "exclude" | "blacklist" ]
#  db      ::= [ "/regexp/" | "string" | "`string`" | "*" ]
#  tbl     ::= [ "/regexp/" | "string" | "`string`" | "*" ]
#  col_val ::= "column_name"
#  tbl     ::= [ "/regexp/" | "string" | "`string`" | "*" ]
#
# See http://maxwells-daemon.io/filtering for more details
#
filter= exclude: *.*, include: market_divide.md_customer_divide_info, include: market_divide.md_customer_divide_info_agent

# javascript filter
# maxwell can run a bit of javascript for each row if you need very custom filtering/data munging.
# See http://maxwells-daemon.io/filtering/#javascript_filters for more details
#
#javascript=/path/to/javascript_filter_file

#       *** encryption ***

# Encryption mode. Possible values are none, data, and all. (default none)
#encrypt=none

# Specify the secret key to be used
#secret_key=RandomInitVector

#       *** monitoring ***

# Maxwell collects metrics via dropwizard. These can be exposed through the
# base logging mechanism (slf4j), JMX, HTTP or pushed to Datadog.
# Options: [jmx, slf4j, http, datadog]
# Supplying multiple is allowed.
#metrics_type=jmx,slf4j

# The prefix maxwell will apply to all metrics
#metrics_prefix=MaxwellMetrics # default MaxwellMetrics

# Enable (dropwizard) JVM metrics, default false
#metrics_jvm=true

# When metrics_type includes slf4j this is the frequency metrics are emitted to the log, in seconds
#metrics_slf4j_interval=60

# When metrics_type includes http or diagnostic is enabled, this is the port the server will bind to.
#http_port=8080

# When metrics_type includes http or diagnostic is enabled, this is the http path prefix, default /.
#http_path_prefix=/some/path/

# ** The following are Datadog specific. **
# When metrics_type includes datadog this is the way metrics will be reported.
# Options: [udp, http]
# Supplying multiple is not allowed.
#metrics_datadog_type=udp

# datadog tags that should be supplied
#metrics_datadog_tags=tag1:value1,tag2:value2

# The frequency metrics are pushed to datadog, in seconds
#metrics_datadog_interval=60

# required if metrics_datadog_type = http
#metrics_datadog_apikey=API_KEY

# required if metrics_datadog_type = udp
#metrics_datadog_host=localhost # default localhost
#metrics_datadog_port=8125 # default 8125

# Maxwell exposes http diagnostic endpoint to check below in parallel:
# 1. binlog replication lag
# 2. producer (currently kafka) lag

# To enable Maxwell diagnostic
#http_diagnostic=true # default false

# Diagnostic check timeout in milliseconds, required if diagnostic = true
#http_diagnostic_timeout=10000 # default 10000

#    *** misc ***

# maxwell's bootstrapping functionality has a couple of modes.
#
# In "async" mode, maxwell will output the replication stream while it
# simultaneously outputs the database to the topic.  Note that it won't
# output replication data for any tables it is currently bootstrapping -- this
# data will be buffered and output after the bootstrap is complete.
#
# In "sync" mode, maxwell stops the replication stream while it
# outputs bootstrap data.
#
# async mode keeps ops live while bootstrapping, but carries the possibility of
# data loss (due to buffering transactions).  sync mode is safer but you
# have to stop replication.
#bootstrapper=async [sync, async, none]

# output filename when using the "file" producer
#output_file=/path/to/file

你可能感兴趣的:(Maxwell)