Dashboarding and alerting
Grafana intro
To display the gathered and stored metrics the pgwatch2 project has decided to rely heavily on the popular Grafana dashboarding solution. This means only though that it’s installed in the default Docker images and there’s a set of predefined dashboards available to cover most of the metrics gathered via the Preset Configs.
This does not mean though that Grafana is in any way tightly coupled with project’s other components - quite the opposite actually, one can use any other means / tools to use the metrics data gathered by the pgwatch2 daemon.
Currently there are around 30 preset dashboards available for PostgreSQL data sources and about 25 for InfluxDB. The reason why InfluxDB (originally the only supported metrics storage DB) has less, is the fact that the custom query language is just much more limiting compared to standard SQL so that it makes it hard to build more complex dashboards. Due to that nowadays, if metric gathering volumes are not a problem, we recommend using Postgres storage for most users.
Note though that most users will probably want to always adjust the built-in dashboards slightly (colors, roundings, etc) , so that they should be taken only as examples to quickly get started. NB! Also note that in case of changes it’s not recommended to change the built-in ones, but use the Save as features - this will allow later to easily update all the dashboards en masse per script, without losing any custom user changes.
Links:
Built-in dashboards for PostgreSQL (TimescaleDB) storage
Built-in dashboards for InfluxDB storage
Alerting
Alerting is very conveniently also supported by Grafana in a simple point-and-click style - see here for the official documentation. In general all more popular notification services are supported and it’s pretty much the easiest way to quickly start with PostgreSQL alerting on a smaller scale. For enterprise usage with hundreds of instances it’s might get too “clicky” though and there are also some limitations - currently you can set alerts only on Graph panels and there must be no variables used in the query so you cannot use most of the pre-created pgwatch2 graphs, but need to create your own.
Nevertheless, alerting via Grafana is s a good option for lighter use cases and there’s also a preset dashboard template named “Alert Template” from the pgwatch2 project to give you some ideas on what to alert on.
Note though that alerting is always a bit of a complex topic - it requires good understanding of PostgreSQL operational metrics and also business criticality background infos, so we don’t want to be too opinionated here and it’s up to the users to implement.