-
Notifications
You must be signed in to change notification settings - Fork 812
Agent 4.0.0 for Linux , Mac OS and FreeBSD
The Datadog Agent 4.0.0 was released on 12-16-2013
This version contains various bug fixes and new features that are mostly impacting the following integrations:
If you are already using these integrations, please read the corresponding sections before upgrading to 4.0.0.
The Datadog Agent 4.0.0 also contains some new features and bug fixes, here is the rest of the Changelog.
-
The new version of the Agent includes a new in-house JMX Connector (JMXFetch: https://github.com/Datadog/jmxfetch) which is more efficient and more reliable than the previous connector (JMXTerm). The new connector will be limited to 350 metrics per instance. This limit was implemented, as depending on your configuration, the check could collect tens of thousands of metrics which would cause high resources consumption (CPU/Memory/Network). To see if you are hitting that limit, you can simply read the info page after the Agent has started.
- If you are hitting the limit, you can easily customize the metrics you want to collect by following this guide:
- If you need to increase the limit of 350 metrics per instance, please get in touch with support@datadoghq.com
-
For these integrations (JMX, Tomcat, Cassandra, Solr, ActiveMQ), support of old-style configuration (using datadog.conf) is removed. Agent 4.0.0 includes an automatic migration script which will create a matching yaml configuration file in your conf.d directory. However, it won't modify your datadog.conf file which means that if you had the old-style configuration in datadog.conf and if you edit one of those checks in this file, it won't have any effect. To edit the configuration of the integrations, edit the matching file in your conf.d directory.
You were then using the previous version of our Cassandra check that was collecting metrics using NodeTool. The new version is collecting metrics directly from the JMX MBeanserver. Your configuration of the check will automatically be migrated to a checks.d, which means it will create a cassandra.yaml file in your conf.d directory. Directly using JMX will make the check more reliable and more efficient. It will also give you more detailed metrics (on a per Column Family/Keyspace basis). However most of your metrics will have a different name. It means that if you had any existing custom dashboards or alerts you will have to update them.
You are already collecting metrics directly from JMX. You won't need to update any dashboard or alerts. However the 100 limits per instance limit will apply (see above).
The latest versions of ActiveMQ included a change in the JMX MBean names, and therefore it requires a different configuration to collect metrics from the proper MBeans:
- To use the ActiveMQ integration for ActiveMQ >= 5.8, you need to use the activemq_58.yaml file which is preconfigured to get metrics from the latests versions of ActiveMQ.
-
The previous version of the integration, bundled with dd-agent 3.x creates artificial hosts for each backend present in the HAProxy instance. A number of customers have reported that this is confusing and we've fixed that (the technical details are here: https://github.com/DataDog/dd-agent/issues/617 ). The new version of the integration won't report the backend metrics as coming from these artificial hosts. There will be another tag to identify them (as explained in the github issue). It means that you will have to update any custom dashboard and metric alerts that you may have created to use the new tags.
-
By default, the new version of the Datadog Agent will only collect metrics from the "aggregate" lines of HAProxy as those are the most useful. If you want to keep collecting metrics from every backend, you need to add the following flag in your haproxy.yaml file:
collect_aggregates_only: False
(See https://github.com/DataDog/dd-agent/blob/master/conf.d/haproxy.yaml.example#L8)
In the previous version of the Datadog Agent, the metric redis.net.commands was wrongly computed.
The new version fixes that which means that you might see a significant change of its value when you upgrade to the new version of the Agent.
See this commit for more information.
- Add an option to forward every packet received to the Dogstatsd server: How to use
- Collect EC2 tags at startup using IAM: (fd404f2c7)
- Add a CouchBase integration #700 (Thanks @jslatts)
- Add a Postfix integration #610 (Thanks @phantasm66)
- Add the connections creation rate in the nginx check #665 (Thanks @dcrosta)
- Add more metrics to the postgres integration for postgres >= 9.2 #683
- Make postgres database name configurable #656
- Add the possibility to monitor more RabbitMQ Queues: (51dd72b)
- Add more metrics to the HAProxy integration regarding response codes: (07f180ae)