The primary stateful data store for litmus's chaos center is a mongoDB statefulset. It powers all the features in chaos center including authentication, chaos injection and analysis, chaos delegate connections etc. Apart from the primary data store, the portal provides a way to add Prometheus (TSDB) data sources for time series data visualization and monitoring. The principle of Open observability drives the chaos center and allows users to manage the data sources for a project in multiple topologies for chaos delegate specific and cross chaos delegate dashboards. The system allows to either have multiple data sources or a single data source scraping metrics and events from all target chaos delegates for monitoring chaos impact on application OR infrastructure.
The following should be required before knowing about managing data sources in chaos center:
Data flow architecture
The chaos center provides the functionality to connect multiple prometheus data sources across projects to harness insights on application or system behavior during chaos manifested on real-time monitoring dashboards. Litmus follows an open observability model which enables users to plug metrics from any Prometheus exporter into the data source connected to a chaos center project to visualize the same on dashboards. Along with system and application metrics the metrics exposed by
chaos-exporter service (a prometheus metrics exporter for chaos injection events and results or verdicts) on the execution plane or target chaos delegate's cluster is necessary to be ingested into the same data source to be connected to the project in order to facilitate chaos interleaving. Existing monitoring infrastructure for observing the target chaos delegate's cluster can also be used if it is prometheus based, plugging the metrics from
chaos-exporter in such cases should be sufficient.
Health check for Time-series database
The query timeout is used for all the queries associated with all the dashboards connected to the given data source, including querying data while editing the dashboard queries, although the default request timeout for the health check of the data source while connecting, updating or listing it is
The scrape interval is used to control the lower limit of minStep for queries multiplying by denominator of query resolution for a dashboard consuming the data source; the same might be used for limiting the refresh rate for dashboard views with relative time range in later versions of the Litmus center.
The Litmus center supports Prometheus 2.1 or later.
LitmusChaos facilitates in-house real-time monitoring for
verdicts metrics exposed by the native
chaos-exporter for each target chaos delegate. These events and metrics can be scraped from a Prometheus TSDB connected to chaos center to overlay on top of application performance and infrastructure monitoring graphs. Some considerations for health-check, metrics scraping interval and version support have been stated.