GridGain provides extensive diagnostic information in a range of industry-standard formats, including JMX, OpenCensus and OpenTracing. This means that it can be integrated with almost any observation platform, though the method of doing so is not always entirely obvious. This repo includes sample configurations and guides for integrating with a number of different systems.
This is by no means a complete list. The absence of a provider should not be construed as GridGain not working well with it. Instead, it means either that we do not come across it on a regular basis, or that integration is trivial and we have never been asked to help.
These guides should not be considered as "production ready" configurations. Rather, they are starting points. The exact metrics that you should monitor and alert on depend on your cluster's workload and the facilities that you're using. Please talk to GridGain to discuss your requirements.
Note that these guides should generally work unchanged with Apache Ignite, though they may not have been tested.
All the systems that GridGain connects to use the same underlying metrics framework. There are some minor differences. For example, OpenCensus is very specifically designed to export metrics, that is, numbers that relate to particular aspect of the system. JMX, on the other hand, is more general and can expose lists of things and strings. However, if it's possible to export them, all metrics are available everywhere.
This means that you won't get "better" metrics by switching from, say, JMX to Open Census. The integration might be more direct or easier, but the metrics will be the same.
A corollary of this is that if you're getting the "wrong" numbers in one tool, switching to another won't help.
The most common gotcha is that some metrics are disabled by default. Make sure the following are enabled in your cluster:
- Cache. Enable metrics by setting the
StatisticsEnabledproperty to true - Data Storage. Enable by setting the
MetricsEnabledproperty to true - Data Region. Enable by setting the
MetricsEnabledproperty to true
Cache metrics can be enabled at runtime. Data region and data storage metrics require a node restart.
GridGain Control Center is a management and monitoring tool for GridGain and Apache Ignite clusters. It features a graphical user interface that helps you perform administrative tasks and monitor your clusters.
GridGain includes native support for Control Center. Sign up for the SaaS version here.
New Relic is an "all-in-one observability" platform and includes the ability to monitor metrics, logs and traces.
Prometheus is a popular monitoring tool supported by the Cloud Native Computing Foundation, the same people as Kubernetes.