This is documentation for the next version of Loki. For the latest stable release, go to the latest version.
Grafana Loki supports the following official clients for sending logs:
There are also a number of third-party clients, see Unofficial clients.
Picking a client
While all clients can be used simultaneously to cover multiple use cases, which client is initially picked to send logs depends on your use case.
Promtail is the client of choice when you’re running Kubernetes, as you can configure it to automatically scrape logs from pods running on the same node that Promtail runs on. Promtail and Prometheus running together in Kubernetes enables powerful debugging: if Prometheus and Promtail use the same labels, users can use tools like Grafana to switch between metrics and logs based on the label set.
Promtail is also the client of choice on bare-metal since it can be configured
to tail logs from all files given a host path. It is the easiest way to send
logs to Loki from plain-text files (e.g., things that log to
Lastly, Promtail works well if you want to extract metrics from logs such as counting the occurrences of a particular message.
Docker Logging Driver
When using Docker and not Kubernetes, the Docker logging driver for Loki should be used as it automatically adds labels appropriate to the running container.
Fluentd and Fluent Bit
The Fluentd and Fluent Bit plugins are ideal when you already have Fluentd deployed
and you already have configured
Fluentd also works well for extracting metrics from logs when using its Prometheus plugin.
If you are already using logstash and/or beats, this will be the easiest way to start. By adding our output plugin you can quickly try Loki without doing big configuration changes.
This is a good choice if you’re looking to try out Loki in a low-footprint way or if you wish to monitor AWS lambda logs in Loki.
Please note that the Loki API is not stable yet, so breaking changes might occur when using or writing a third-party client.