如何在AWS (ElasticMapReduce)中监视Apache？

Question

我现在已经设置了Flink，并在EMR上运行了一个作业，现在我正试图通过将指标发送到prometheus来添加监视。

在EMR上运行Flink遇到了一个问题。我使用Terraform来提供EMR (我在下载和运行作业之后运行ansible )。打开盒子，它看起来不像EMR的Flink发行版包括可选的jars (flink-指标-prometheus，flink-cep等等)。

看看弗林克的文档，上面写着

“要使用此记者，必须将/opt/flink-metrics-prometheus-1.6.1.jar复制到Flink发行版的/lib文件夹”https://ci.apache.org/projects/flink/flink-docs-release-1.6/monitoring/metrics.html#prometheuspushgateway-orgapacheflinkmetricsprometheusprometheuspushgatewayreporter

但是，当登录到EMR主节点时，/etc/flink或/usr/lib/flink都没有一个名为opts的目录，而且我在任何地方都看不到flink-metrics-prometheus-1.6.1.jar。

我知道Flink还有其他可选的库，如果您想要使用flink-cep，通常需要复制它们，但是我不知道如何在使用EMR时做到这一点。

这是我得到的例外，我认为这是因为它无法在其类路径中找到度量jar。

java.lang.ClassNotFoundException: org.apache.flink.metrics.prometheus.PrometheusPushGatewayReporter
    at java.net.URLClassLoader.findClass(URLClassLoader.java:382)
    at java.lang.ClassLoader.loadClass(ClassLoader.java:424)
    at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:349)
    at java.lang.ClassLoader.loadClass(ClassLoader.java:357)
    at java.lang.Class.forName0(Native Method)
    at java.lang.Class.forName(Class.java:264)
    at org.apache.flink.runtime.metrics.MetricRegistryImpl.<init>(MetricRegistryImpl.java:144)
    at org.apache.flink.runtime.entrypoint.ClusterEntrypoint.createMetricRegistry(ClusterEntrypoint.java:419)
    at org.apache.flink.runtime.entrypoint.ClusterEntrypoint.initializeServices(ClusterEntrypoint.java:276)
    at org.apache.flink.runtime.entrypoint.ClusterEntrypoint.runCluster(ClusterEntrypoint.java:227)
    at org.apache.flink.runtime.entrypoint.ClusterEntrypoint.lambda$startCluster$0(ClusterEntrypoint.java:191)
    at java.security.AccessController.doPrivileged(Native Method)
    at javax.security.auth.Subject.doAs(Subject.java:422)
    at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1844)
    at org.apache.flink.runtime.security.HadoopSecurityContext.runSecured(HadoopSecurityContext.java:41)
    at org.apache.flink.runtime.entrypoint.ClusterEntrypoint.startCluster(ClusterEntrypoint.java:190)
    at org.apache.flink.yarn.entrypoint.YarnSessionClusterEntrypoint.main(YarnSessionClusterEntrypoint.java:137)

地形中的EMR资源

  resource "aws_emr_cluster" "emr_flink" {
  name          = "ce-emr-flink-arn"
  release_label = "emr-5.20.0" # 5.21.0 is not found, could be a region thing
  applications  = ["Flink"]

  ec2_attributes {
    key_name                          = "ce_test"
    subnet_id                         = "${aws_subnet.ce_test_subnet_public.id}"
    instance_profile                  = "${aws_iam_instance_profile.emr_profile.arn}"
    emr_managed_master_security_group = "${aws_security_group.allow_all_vpc.id}"
    emr_managed_slave_security_group  = "${aws_security_group.allow_all_vpc.id}"
    additional_master_security_groups  = "${aws_security_group.external_connectivity.id}"
    additional_slave_security_groups  = "${aws_security_group.external_connectivity.id}"
  }

  ebs_root_volume_size = 100
  master_instance_type = "m4.xlarge"
  core_instance_type   = "m4.xlarge"
  core_instance_count  = 2

  service_role = "${aws_iam_role.iam_emr_service_role.arn}"

  configurations_json = <<EOF
[
  {
    "Classification": "flink-conf",
    "Properties": {
        "parallelism.default": "8",
        "state.backend": "RocksDB",
        "state.backend.async": "true",
        "state.backend.incremental": "true",
        "state.savepoints.dir": "file:///savepoints",
        "state.checkpoints.dir": "file:///checkpoints",
        "web.submit.enable": "true",
        "metrics.reporter.promgateway.class": "org.apache.flink.metrics.prometheus.PrometheusPushGatewayReporter",
        "metrics.reporter.promgateway.host": "${aws_instance.monitoring.private_ip}",
        "metrics.reporter.promgateway.port": "9091",
        "metrics.reporter.promgateway.jobName": "ce-test",
        "metrics.reporter.promgateway.randomJobNameSuffix": "true",
        "metrics.reporter.promgateway.deleteOnShutdown": "false"
    }
  }
]
EOF
}

我怀疑我可能不得不在引导阶段下载Jar，但我想先检查一下，看看是否有这样做的例子

Answer 1

我还没有使用Terraform，但请注意，您通常需要在EMR中为主从双方提供(设置jars)。找出EMR认为jars应该去哪里的一种方法是，在作业运行时登录到一个从站，执行ps auxwww | grep java，查找TaskManager进程，查看启动时添加到类路径中的jars，并找到这些jars位于服务器上的位置。或者至少在过去对我起过作用。

Answer 2

我选择了EMR释放的emr-5.24.0，我用suceed的进水数据库.jar进行监测。

我已经将.jar文件复制到/usr/lib/flink/lib文件夹，并使用以下bash命令重新启动Flink集群(具有sudo权限)。

/usr/lib/flink/bin/stop-cluster.sh && /usr/lib/flink/bin/stop-cluster.sh

我想你可以用普罗米修斯同样的步骤来解决你的问题

[ec2-user@ip-10-0-11-17 ~]$ cd /usr/lib/flink/opt/flink-metrics-
flink-metrics-datadog-1.8.0.jar     flink-metrics-influxdb-1.8.0.jar    flink-metrics-slf4j-1.8.0.jar
flink-metrics-graphite-1.8.0.jar    flink-metrics-prometheus-1.8.0.jar  flink-metrics-statsd-1.8.0.jar


[ec2-user@ip-10-0-11-17 ~]$ ll /usr/lib/flink/opt/flink-metrics-prometheus-1.8.0.jar
-rw-r--r-- 1 root root 101984 may 14 19:21 /usr/lib/flink/opt/flink-metrics-prometheus-1.8.0.jar


[ec2-user@ip-10-0-11-17 ~]$ uname -a
Linux ip-10-0-11-17 4.14.114-83.126.amzn1.x86_64 #1 SMP Tue May 7 02:26:58 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux