4.0 KiB
Introduction
The StackLight Collector Plugin is used to install and configure several software components that are used to collect and process all the data that we think is relevant to provide deep operational insights about your OpenStack environment. These finely integrated components are collectively referred to as the StackLight Collector (or just the Collector).
Note
The Collector has evolved over time and so the term 'collector' is a little bit of a misnomer since it is more of a smart monitoring agent than a mere data 'collector'.
The Collecor is a key component of the so-called Logging, Monitoring and Alerting toolchain of Mirantis OpenStack (a.k.a StackLight).
The Collector is installed on every node of your OpenStack environment. Each Collector is individually responsible for supporting all the monitoring functions of your OpenStack environment for both the operating system and the services running on the node. Note also that the Collector running on the primary controller (the controller which owns the management VIP) is called the Aggregator since it performs additional aggregation and correlation functions. The Aggregator is the central point of convergence for all the faults and anomalies detected at the node level. The fundamental role of the Aggregator is to issue an opinion about the health status of your OpenStack environment at the cluster level. As such, the Collector may be viewed as a monitoring agent for cloud infrastructure clusters.
The main building blocks of the Collector are:
- collectd which comes bundled with a collection of monitoring plugins. Some of them are standard collectd plugins while others are purpose-built plugins written in python to perform various OpenStack services checks.
- Heka, a golang data processing swiss army knife by Mozilla. Heka supports a number of standard input and output plugins that allows to ingest data from a variety of sources including collectd, log files and RabbitMQ, as well as to persist the operational data to external backend servers like Elasticsearch, InfluxDB and Nagios for search and further processing.
- A collection of Heka plugins written in Lua which does the actual data processing such as running metrics transformations, running alarms and logs parsing.
Note
An important function of the Collector is to normalize the operational data into an internal Heka message structure representation that can be ingested into the Heka's stream processing pipeline. The stream processing pipeline uses matching policies to route the Heka messages to the Lua plugins that will perform the actual data computation functions.
There are three types of Lua plugins that were developed for the Collector:
- The decoder plugins to sanitize and normalize the ingested data.
- The filter plugins to process the data.
- The encoder plugins to serialize the data that is sent to the backend servers.
There are five types of data sent by the Collector (and the Aggregator) to the backend servers:
- The logs and the notifications, which are referred to as events, sent to Elasticsearch for indexing.
- The metric's time-series sent to InfluxDB.
- The annotation sent to InfluxDB.
- The OpenStack environment clusters health status sent as passive checks to Nagios
Note
The annotations are like notification messages which are exposed in Grafana. They contain information about the anomalies and faults that have been detected by the Collector. They basicaly contain the same information as the passive checks sent to Nagios. In addition, they may contain 'hints' about what the Collector think could be the root cause of a problem.