DataBot

Internal

• projects
• events

Overview

A low-overhead O/S level event collector that generates events-compatible events. It is designed to run in background, collect times events and channel them (to files, over the network, etc.) It is capable of collecting memory, CPU, etc. usage statistics, as well as WildFly management domain model and JMX metrics.

os-stats User Manual

In-line help is available with:

os-stats --help

os-stats Metric Reference

os-stats Metric Reference

os-stats Development

GitHub

git clone git@github.com:NovaOrdis/os-stats.git

os-stats TODO

• Some OS metrics can be obtained from the JVM:

      conf:
        - include:
            domain: java.lang
            bean:
               - java.lang:type=OperatingSystem
               - java.lang:type=Threading
            attribute:
               OpenFileDescriptorCount:
                 alias: jmx.OpenFileDescriptorCount
                 metric_type: counter
               MaxFileDescriptorCount:
                 alias: jmx.MaxFileDescriptorCount
                 metric_type: counter
               TotalStartedThreadCount:
                 alias: jmx.TotalStartedThreadCount
                 metric_type: counter
               PeakThreadCount:
                 alias: jmx.PeakThreadCount
                 metric_type: counter

• 10/15/16 Does not work with JDG 7:

@Timer-0 07:14:05,153 WARN  [DataCollectionTimerTask] failed to collect data: org.jboss.as.cli.CommandContextFactory.newCommandContext(Ljava/lang/String;ILjava/lang/String;[CZZI)Lorg/jboss/as/cli/CommandContext;

• 10/04/16 os-stats failed to gather correct CPU consumption data from top. It seems to read a constant value. Troubleshoot and fix. Possibly calculate the values the same way top does, by reading from /proc files.

• Explore using iftop to gather up/down bandwidth usage for each host:port pair.

• JDG testing
  • Network bandwidth statistics, integrate iftop.
  • Access to MBeans. Paul's e-mail The mbean of the transport for the bridge channel exposes the attributes: num_bytes_sent and num_bytes_received.

• For four CLI metrics (2 for one instance and 2 for another, I get 4 cli.xml warnings. Normally I should get only 2 because I should share the clients.

[fsgapp@ln001xsjdg0003 etc]$ WARN: can't find jboss-cli.xml. Using default configuration values.
@Timer-0 17:28:55,620 INFO  [xnio] XNIO Version 3.0.14.GA-redhat-1
@Timer-0 17:28:55,623 INFO  [nio] XNIO NIO Implementation Version 3.0.14.GA-redhat-1
@Timer-0 17:28:55,634 INFO  [remoting] JBoss Remoting version 3.3.5.Final-redhat-1
WARN: can't find jboss-cli.xml. Using default configuration values.
WARN: can't find jboss-cli.xml. Using default configuration values.
WARN: can't find jboss-cli.xml. Using default configuration values.

• Fix logging, get rid of:

NOMBP2:tmp ovidiu$ @Timer-0 18:37:22,052 INFO  [xnio] XNIO Version 3.0.15.GA-redhat-1
@Timer-0 18:37:22,058 INFO  [nio] XNIO NIO Implementation Version 3.0.15.GA-redhat-1
@Timer-0 18:37:22,080 INFO  [remoting] JBoss Remoting version 3.3.7.Final-redhat-1

• Disconnect the JBossCliMetricSource instances on os-stats shutdown.
• Support for password externalization. Then update Os-stats_Metric_Reference#JBoss_CLI_Metrics with details.
• Warn if the collection time takes so long that it interferes with sampling interval length. Tests.
• Implement os-stats --help metrics that introspects and displays the list of built-in metrics.
• Current OSBase.execute() implementation is naive, implement it correctly.
• Integrate the functionality and get rid of https://github.com/NovaOrdis/shell-tools/blob/master/jboss-os-memory-stats.sh