Before you can react to a change in the system you need to know that something has changed.
Replicante’s first task is to observe all nodes and generate events to reflect changes to nodes and clusters. These events are internally used to drive features but are also recoded for users to see.
Having access to historical events can provide valuable insight:
Being able to correlate datastore events with application errors, performance issues, or other unusual activity is key into improving your services and products.
The simplest way to view the events history is to check out the
Events page in the WebUI.
Observed and generated events are emitted onto a stream.
Events order is guaranteed for events in the same cluster but events from different clusters can be interleaved in any order. System events order is also guaranteed with respect to other system events but not to cluster events.
The stream of events acts as the source of information for most of the system and drives the actions subsystem. Both causes and effects of system state changes will be reflected in this stream.
On top of being a foundation for Replicante itself, the use of an events stream means that external systems can be designed to follow (or subscribe to) the stream to be notified of events as they occur.
Such powerful extension point opens the system up for almost endless integrations. To name just a few:
Replicante periodically emits a special kind of event to describe the current state of an entity.
Since Replicante monitors distributed systems it is not possible to consistently provide a point in time view of monitored clusters. It is on the other hand possible to observe individual parts of the cluster state consistently.
SNAPSHOT_* events each describe a different part of the system at a different point in time.
External systems can make use of these snapshot events to synchronise their state with the observed
state of the cluster and compensate for missing events or other errors.
Event snapshots can be disabled by setting the
events.snapshots.enabled option to
This feature is enabled by default.
events.snapshots.frequencyconfiguration options controls how often snapshots should be emitted.
The system does not currently guarantee an upper bound to the time between intervals.
This feature requires the Grafana component to be enabled enabled
components.grafana configuration option; enabled by default).
A set of endpoints compatible with the
simple JSON datasource
is available under the
The annotations integration mainly provide a specialised query interface to stored events so most of the configuration is on the Grafana side:
Simple JSONdatasource to point to
http://REPLICANTE_HOST:REPLICANTE_PORT/api/unstable/grafanaif not already available.
By default, the first 1000 non-
SNAPSHOT_* events in the queried interval are returned.
The following filters can be used to change what is returned:
cluster_id: (string) filter events by originating cluster ID.
event: (string) filter events by event type.
false) exclude events not originating from a cluster.
limit: (integer) maximum number of events to return.