Ganglia 3.6.0 监控 Hadoop 2.2.0

在《Ganglia3.6.0 安装步骤 (含python module)》一文中，安装完毕以后，ganglia可以正常监控到机器负载、cpu、磁盘等机器相关信息。但是，如果需要进一步通过ganglia监控hadoop集群，则需要配置hadoop metrics配置文件及适量修改ganglia配置。

一、hadoop相关配置

HADOOP_PATH/etc/hadoop/目录下有两个配置文件：hadoop-metrics.properties和hadoop-metrics2.properties。

1. hadoop-metrics.properties

用于hadoop与3.1版本以前的ganglia集成做监控的配置文件（在ganglia3.0到3.1的过程中，消息的格式发生了重要的变化，不兼容之前的版本）。

示例如下：

# Configuration of the "dfs" context for null
##dfs.class=org.apache.hadoop.metrics.spi.NullContext

# Configuration of the "dfs" context for file
#dfs.class=org.apache.hadoop.metrics.file.FileContext
#dfs.period=10
#dfs.fileName=/tmp/dfsmetrics.log

# Configuration of the "dfs" context for ganglia
# Pick one: Ganglia 3.0 (former) or Ganglia 3.1 (latter)
# dfs.class=org.apache.hadoop.metrics.ganglia.GangliaContext
dfs.class=org.apache.hadoop.metrics.ganglia.GangliaContext31
dfs.period=10
dfs.servers=Mas2:8649

# Configuration of the "mapred" context for null
##mapred.class=org.apache.hadoop.metrics.spi.NullContext

# Configuration of the "mapred" context for file
#mapred.class=org.apache.hadoop.metrics.file.FileContext
#mapred.period=10
#mapred.fileName=/tmp/mrmetrics.log

# Configuration of the "mapred" context for ganglia
# Pick one: Ganglia 3.0 (former) or Ganglia 3.1 (latter)
# mapred.class=org.apache.hadoop.metrics.ganglia.GangliaContext
mapred.class=org.apache.hadoop.metrics.ganglia.GangliaContext31
mapred.period=10
mapred.servers=Mas2:8649

# Configuration of the "jvm" context for null
#jvm.class=org.apache.hadoop.metrics.spi.NullContext

# Configuration of the "jvm" context for file
#jvm.class=org.apache.hadoop.metrics.file.FileContext
#jvm.period=10
#jvm.fileName=/tmp/jvmmetrics.log

# Configuration of the "jvm" context for ganglia
# jvm.class=org.apache.hadoop.metrics.ganglia.GangliaContext
jvm.class=org.apache.hadoop.metrics.ganglia.GangliaContext31
jvm.period=10
jvm.servers=Mas2:8649

# Configuration of the "rpc" context for null
##rpc.class=org.apache.hadoop.metrics.spi.NullContext

# Configuration of the "rpc" context for file
#rpc.class=org.apache.hadoop.metrics.file.FileContext
#rpc.period=10
#rpc.fileName=/tmp/rpcmetrics.log

# Configuration of the "rpc" context for ganglia
# rpc.class=org.apache.hadoop.metrics.ganglia.GangliaContext
rpc.class=org.apache.hadoop.metrics.ganglia.GangliaContext31
rpc.period=10
rpc.servers=Mas2:8649

# Configuration of the "ugi" context for null
##ugi.class=org.apache.hadoop.metrics.spi.NullContext

# Configuration of the "ugi" context for file
#ugi.class=org.apache.hadoop.metrics.file.FileContext
#ugi.period=10
#ugi.fileName=/tmp/ugimetrics.log

# Configuration of the "ugi" context for ganglia
# ugi.class=org.apache.hadoop.metrics.ganglia.GangliaContext
ugi.class=org.apache.hadoop.metrics.ganglia.GangliaContext31
ugi.period=10
ugi.servers=Mas2:8649

2. hadoop-metrics2.properties

用于hadoop与3.1版本以后的ganglia集成做监控的配置文件（本文采用此配置文件）。

示例如下（根据具体的namenode、datanode或是其他，选择相应的配置）：

#
#   Licensed to the Apache Software Foundation (ASF) under one or more
#   contributor license agreements.  See the NOTICE file distributed with
#   this work for additional information regarding copyright ownership.
#   The ASF licenses this file to You under the Apache License, Version 2.0
#   (the "License"); you may not use this file except in compliance with
#   the License.  You may obtain a copy of the License at
#
#       http://www.apache.org/licenses/LICENSE-2.0 #
#   Unless required by applicable law or agreed to in writing, software
#   distributed under the License is distributed on an "AS IS" BASIS,
#   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
#   See the License for the specific language governing permissions and
#   limitations under the License.
#

# syntax: [prefix].[source|sink].[instance].[options]
# See javadoc of package-info.java for org.apache.hadoop.metrics2 for details

*.sink.file.class=org.apache.hadoop.metrics2.sink.FileSink
# default sampling period, in seconds
*.period=10

# The namenode-metrics.out will contain metrics from all context
#namenode.sink.file.filename=namenode-metrics.out
# Specifying a special sampling period for namenode:
#namenode.sink.*.period=8

#datanode.sink.file.filename=datanode-metrics.out

# the following example split metrics of different
# context to different sinks (in this case files)
#jobtracker.sink.file_jvm.context=jvm
#jobtracker.sink.file_jvm.filename=jobtracker-jvm-metrics.out
#jobtracker.sink.file_mapred.context=mapred
#jobtracker.sink.file_mapred.filename=jobtracker-mapred-metrics.out

#tasktracker.sink.file.filename=tasktracker-metrics.out

#maptask.sink.file.filename=maptask-metrics.out

#reducetask.sink.file.filename=reducetask-metrics.out

##
*.sink.ganglia.class=org.apache.hadoop.metrics2.sink.ganglia.GangliaSink31
*.sink.ganglia.period=10

*.sink.ganglia.slope=jvm.metrics.gcCount=zero,jvm.metrics.memHeapUsedM=both
*.sink.ganglia.dmax=jvm.metrics.threadsBlocked=70,jvm.metrics.memHeapUsedM=40

#namenode.sink.ganglia.servers=Mas2:8649
#resourcemanager.sink.ganglia.servers=namenode1:8649

datanode.sink.ganglia.servers=namenode1:8649
#nodemanager.sink.ganglia.servers=namenode1:8649

#maptask.sink.ganglia.servers=namenode1:8649
#reducetask.sink.ganglia.servers=namenode1:8649

二、ganglia相关配置

1. gmeted.conf

用于收集各节点上传上来的监控数据。

示例如下（启用示例文件中开启的配置选项即可）：

# This is an example of a Ganglia Meta Daemon configuration file
#                http://ganglia.sourceforge.net/ #
#
#-------------------------------------------------------------------------------
# Setting the debug_level to 1 will keep daemon in the forground and
# show only error messages. Setting this value higher than 1 will make
# gmetad output debugging information and stay in the foreground.
# default: 0
# debug_level 10
#
#-------------------------------------------------------------------------------
# What to monitor. The most important section of this file.
#
# The data_source tag specifies either a cluster or a grid to
# monitor. If we detect the source is a cluster, we will maintain a complete
# set of RRD databases for it, which can be used to create historical
# graphs of the metrics. If the source is a grid (it comes from another gmetad),
# we will only maintain summary RRDs for it.
#
# Format:
# data_source "my cluster" [polling interval] address1:port addreses2:port ...
#
# The keyword 'data_source' must immediately be followed by a unique
# string which identifies the source, then an optional polling interval in
# seconds. The source will be polled at this interval on average.
# If the polling interval is omitted, 15sec is asssumed.
#
# If you choose to set the polling interval to something other than the default,
# note that the web frontend determines a host as down if its TN value is less
# than 4 * TMAX (20sec by default).  Therefore, if you set the polling interval
# to something around or greater than 80sec, this will cause the frontend to
# incorrectly display hosts as down even though they are not.
#
# A list of machines which service the data source follows, in the
# format ip:port, or name:port. If a port is not specified then 8649
# (the default gmond port) is assumed.
# default: There is no default value
#
# data_source "my cluster" 10 localhost  my.machine.edu:8649  1.2.3.5:8655
# data_source "my grid" 50 1.3.4.7:8655 grid.org:8651 grid-backup.org:8651
# data_source "another source" 1.3.4.7:8655  1.3.4.8

#data_source "my cluster" localhost
data_source "my cluster" 10 Mas1:8650 Mas2:8650 Sla1:8650 Sla2:8650

#
# Round-Robin Archives
# You can specify custom Round-Robin archives here (defaults are listed below)
#
# Old Default RRA: Keep 1 hour of metrics at 15 second resolution. 1 day at 6 minute
# RRAs "RRA:AVERAGE:0.5:1:244" "RRA:AVERAGE:0.5:24:244" "RRA:AVERAGE:0.5:168:244" "RRA:AVERAGE:0.5:672:244" \
#      "RRA:AVERAGE:0.5:5760:374"
# New Default RRA
# Keep 5856 data points at 15 second resolution assuming 15 second (default) polling. That's 1 day
# Two weeks of data points at 1 minute resolution (average)
#RRAs "RRA:AVERAGE:0.5:1:5856" "RRA:AVERAGE:0.5:4:20160" "RRA:AVERAGE:0.5:40:52704"

#
#-------------------------------------------------------------------------------
# Scalability mode. If on, we summarize over downstream grids, and respect
# authority tags. If off, we take on 2.5.0-era behavior: we do not wrap our output
# in <GRID></GRID> tags, we ignore all <GRID> tags we see, and always assume
# we are the "authority" on data source feeds. This approach does not scale to
# large groups of clusters, but is provided for backwards compatibility.
# default: on
# scalable off
#
#-------------------------------------------------------------------------------
# The name of this Grid. All the data sources above will be wrapped in a GRID
# tag with this name.
# default: unspecified
# gridname "MyGrid"
#
#-------------------------------------------------------------------------------
# The authority URL for this grid. Used by other gmetads to locate graphs
# for our data sources. Generally points to a ganglia/
# website on this machine.
# default: "http://hostname/ganglia/",
#   where hostname is the name of this machine, as defined by gethostname().
# authority "http://mycluster.org/newprefix/"
#
#-------------------------------------------------------------------------------
# List of machines this gmetad will share XML with. Localhost
# is always trusted.
# default: There is no default value
# trusted_hosts 127.0.0.1 169.229.50.165 my.gmetad.org
#
#-------------------------------------------------------------------------------
# If you want any host which connects to the gmetad XML to receive
# data, then set this value to "on"
# default: off
# all_trusted on
#
#-------------------------------------------------------------------------------
# If you don't want gmetad to setuid then set this to off
# default: on
# setuid off
#
#-------------------------------------------------------------------------------
# User gmetad will setuid to (defaults to "nobody")
# default: "nobody"
# setuid_username "nobody"
#
#-------------------------------------------------------------------------------
# Umask to apply to created rrd files and grid directory structure
# default: 0 (files are public)
# umask 022
#
#-------------------------------------------------------------------------------
# The port gmetad will answer requests for XML
# default: 8651
xml_port 8651
#
#-------------------------------------------------------------------------------
# The port gmetad will answer queries for XML. This facility allows
# simple subtree and summation views of the XML tree.
# default: 8652
interactive_port 8652
#
#-------------------------------------------------------------------------------
# The number of threads answering XML requests
# default: 4
# server_threads 10
#
#-------------------------------------------------------------------------------
# Where gmetad stores its round-robin databases
# default: "/var/lib/ganglia/rrds"
rrd_rootdir "/var/www/html/rrds"
#
#-------------------------------------------------------------------------------
# List of metric prefixes this gmetad will not summarize at cluster or grid level.
# default: There is no default value
# unsummarized_metrics diskstat CPU
#
#-------------------------------------------------------------------------------
# In earlier versions of gmetad, hostnames were handled in a case
# sensitive manner
# If your hostname directories have been renamed to lower case,
# set this option to 0 to disable backward compatibility.
# From version 3.2, backwards compatibility will be disabled by default.
# default: 1   (for gmetad < 3.2)
# default: 0   (for gmetad >= 3.2)
case_sensitive_hostnames 0

#-------------------------------------------------------------------------------
# It is now possible to export all the metrics collected by gmetad directly to
# graphite by setting the following attributes.
#
# The hostname or IP address of the Graphite server
# default: unspecified
# carbon_server "my.graphite.box"
#
# The port and protocol on which Graphite is listening
# default: 2003
# carbon_port 2003
#
# default: tcp
# carbon_protocol udp
#
# **Deprecated in favor of graphite_path** A prefix to prepend to the
# metric names exported by gmetad. Graphite uses dot-
# separated paths to organize and refer to metrics.
# default: unspecified
# graphite_prefix "datacenter1.gmetad"
#
# A user-definable graphite path. Graphite uses dot-
# separated paths to organize and refer to metrics.
# For reverse compatibility graphite_prefix will be prepended to this
# path, but this behavior should be considered deprecated.
# This path may include 3 variables that will be replaced accordingly:
# %s -> source (cluster name)
# %h -> host (host name)
# %m -> metric (metric name)
# default: graphite_prefix.%s.%h.%m
# graphite_path "datacenter1.gmetad.%s.%h.%m

# Number of milliseconds gmetad will wait for a response from the graphite server
# default: 500
# carbon_timeout 500

#-------------------------------------------------------------------------------
# Memcached configuration (if it has been compiled in)
# Format documentation at http://docs.libmemcached.org/libmemcached_configuration.html # default: ""
# memcached_parameters "--SERVER=127.0.0.1"
#

2. gmond.conf

用于指定本机监控信息的发送，本机所属的集群等的相关配置（gmond可以指定多播，在本示例中采用将监控数据发送到指定地址）。

示例如下（需要改动的部分为第29~72行）：

/* This configuration is as close to 2.5.x default behavior as possible
The values closely match ./gmond/metric.h definitions in 2.5.x */
globals {
daemonize = yes
setuid = yes
user = nobody
debug_level = 0
max_udp_msg_len = 1472
mute = no
deaf = no
allow_extra_data = yes
host_dmax = 86400 /*secs. Expires (removes from web interface) hosts in 1 day */
host_tmax = 20 /*secs */
cleanup_threshold = 300 /*secs */
gexec = no
# By default gmond will use reverse DNS resolution when displaying your hostname
# Uncommeting following value will override that value.
# override_hostname = "mywebserver.domain.com"
# If you are not using multicast this value should be set to something other than 0.
# Otherwise if you restart aggregator gmond you will get empty graphs. 60 seconds is reasonable
send_metadata_interval = 0 /*secs */

}

/*
* The cluster attributes specified will be used as part of the <CLUSTER>
* tag that will wrap all hosts collected by this instance.
*/
cluster {
name = "my cluster"
owner = "nobody"
latlong = "unspecified"
url = "unspecified"
}

/* The host section describes attributes of the host, like the location */
host {
location = "unspecified"
}

/* Feel free to specify as many udp_send_channels as you like.  Gmond
used to only support having a single channel */
udp_send_channel {
bind_hostname = yes # Highly recommended, soon to be default.
# This option tells gmond to use a source address
# that resolves to the machine's hostname.  Without
# this, the metrics may appear to come from any
# interface and the DNS names associated with
# those IPs will be used to create the RRDs.
#  mcast_join = 239.2.11.71
port = 8649
ttl = 1
}

/* You can specify as many udp_recv_channels as you like as well. */
udp_recv_channel {
#  mcast_join = 239.2.11.71
port = 8649
#  bind = 239.2.11.71
#  retry_bind = true
# Size of the UDP buffer. If you are handling lots of metrics you really
# should bump it up to e.g. 10MB or even higher.
# buffer = 10485760
}

/* You can specify as many tcp_accept_channels as you like to share
an xml description of the state of the cluster */
tcp_accept_channel {
port = 8650
# If you want to gzip XML output
gzip_output = no
}

/* Channel to receive sFlow datagrams */
#udp_recv_channel {
#  port = 6343
#}

/* Optional sFlow settings */
#sflow {
# udp_port = 6343
# accept_vm_metrics = yes
# accept_jvm_metrics = yes
# multiple_jvm_instances = no
# accept_http_metrics = yes
# multiple_http_instances = no
# accept_memcache_metrics = yes
# multiple_memcache_instances = no
#}

/* Each metrics module that is referenced by gmond must be specified and
loaded. If the module has been statically linked with gmond, it does
not require a load path. However all dynamically loadable modules must
include a load path. */
modules {
module {
name = "core_metrics"
}
module {
name = "cpu_module"
path = "modcpu.so"
}
module {
name = "disk_module"
path = "moddisk.so"
}
module {
name = "load_module"
path = "modload.so"
}
module {
name = "mem_module"
path = "modmem.so"
}
module {
name = "net_module"
path = "modnet.so"
}
module {
name = "proc_module"
path = "modproc.so"
}
module {
name = "sys_module"
path = "modsys.so"
}
}

/* The old internal 2.5.x metric array has been replaced by the following
collection_group directives.  What follows is the default behavior for
collecting and sending metrics that is as close to 2.5.x behavior as
possible. */

/* This collection group will cause a heartbeat (or beacon) to be sent every
20 seconds.  In the heartbeat is the GMOND_STARTED data which expresses
the age of the running gmond. */
collection_group {
collect_once = yes
time_threshold = 20
metric {
name = "heartbeat"
}
}

/* This collection group will send general info about this host every
1200 secs.
This information doesn't change between reboots and is only collected
once. */
collection_group {
collect_once = yes
time_threshold = 1200
metric {
name = "cpu_num"
title = "CPU Count"
}
metric {
name = "cpu_speed"
title = "CPU Speed"
}
metric {
name = "mem_total"
title = "Memory Total"
}
/* Should this be here? Swap can be added/removed between reboots. */
metric {
name = "swap_total"
title = "Swap Space Total"
}
metric {
name = "boottime"
title = "Last Boot Time"
}
metric {
name = "machine_type"
title = "Machine Type"
}
metric {
name = "os_name"
title = "Operating System"
}
metric {
name = "os_release"
title = "Operating System Release"
}
metric {
name = "location"
title = "Location"
}
}

/* This collection group will send the status of gexecd for this host
every 300 secs.*/
/* Unlike 2.5.x the default behavior is to report gexecd OFF. */
collection_group {
collect_once = yes
time_threshold = 300
metric {
name = "gexec"
title = "Gexec Status"
}
}

/* This collection group will collect the CPU status info every 20 secs.
The time threshold is set to 90 seconds.  In honesty, this
time_threshold could be set significantly higher to reduce
unneccessary  network chatter. */
collection_group {
collect_every = 20
time_threshold = 90
/* CPU status */
metric {
name = "cpu_user"
value_threshold = "1.0"
title = "CPU User"
}
metric {
name = "cpu_system"
value_threshold = "1.0"
title = "CPU System"
}
metric {
name = "cpu_idle"
value_threshold = "5.0"
title = "CPU Idle"
}
metric {
name = "cpu_nice"
value_threshold = "1.0"
title = "CPU Nice"
}
metric {
name = "cpu_aidle"
value_threshold = "5.0"
title = "CPU aidle"
}
metric {
name = "cpu_wio"
value_threshold = "1.0"
title = "CPU wio"
}
metric {
name = "cpu_steal"
value_threshold = "1.0"
title = "CPU steal"
}
/* The next two metrics are optional if you want more detail...
... since they are accounted for in cpu_system.
metric {
name = "cpu_intr"
value_threshold = "1.0"
title = "CPU intr"
}
metric {
name = "cpu_sintr"
value_threshold = "1.0"
title = "CPU sintr"
}
*/
}

collection_group {
collect_every = 20
time_threshold = 90
/* Load Averages */
metric {
name = "load_one"
value_threshold = "1.0"
title = "One Minute Load Average"
}
metric {
name = "load_five"
value_threshold = "1.0"
title = "Five Minute Load Average"
}
metric {
name = "load_fifteen"
value_threshold = "1.0"
title = "Fifteen Minute Load Average"
}
}

/* This group collects the number of running and total processes */
collection_group {
collect_every = 80
time_threshold = 950
metric {
name = "proc_run"
value_threshold = "1.0"
title = "Total Running Processes"
}
metric {
name = "proc_total"
value_threshold = "1.0"
title = "Total Processes"
}
}

/* This collection group grabs the volatile memory metrics every 40 secs and
sends them at least every 180 secs.  This time_threshold can be increased
significantly to reduce unneeded network traffic. */
collection_group {
collect_every = 40
time_threshold = 180
metric {
name = "mem_free"
value_threshold = "1024.0"
title = "Free Memory"
}
metric {
name = "mem_shared"
value_threshold = "1024.0"
title = "Shared Memory"
}
metric {
name = "mem_buffers"
value_threshold = "1024.0"
title = "Memory Buffers"
}
metric {
name = "mem_cached"
value_threshold = "1024.0"
title = "Cached Memory"
}
metric {
name = "swap_free"
value_threshold = "1024.0"
title = "Free Swap Space"
}
}

collection_group {
collect_every = 40
time_threshold = 300
metric {
name = "bytes_out"
value_threshold = 4096
title = "Bytes Sent"
}
metric {
name = "bytes_in"
value_threshold = 4096
title = "Bytes Received"
}
metric {
name = "pkts_in"
value_threshold = 256
title = "Packets Received"
}
metric {
name = "pkts_out"
value_threshold = 256
title = "Packets Sent"
}
}

/* Different than 2.5.x default since the old config made no sense */
collection_group {
collect_every = 1800
time_threshold = 3600
metric {
name = "disk_total"
value_threshold = 1.0
title = "Total Disk Space"
}
}

collection_group {
collect_every = 40
time_threshold = 180
metric {
name = "disk_free"
value_threshold = 1.0
title = "Disk Space Available"
}
metric {
name = "part_max_used"
value_threshold = 1.0
title = "Maximum Disk Space Used"
}
}

include ("/usr/local/ganglia/etc/conf.d/*.conf")

三、总结

自此，hadoop和ganglia的配置部分完成，接下来重启hadoop和ganglia进程，即可实现监控！