In this guide we show how to configure a Fluentd instance to securely send logs to your Opstrace instance.
We appreciate Fluentd for its robustness, especially with respect to log file rotation handling, node-local log data buffer management, flushing parameterization and its mature error handling architecture.
It is worth mentioning that Fluentd has decent debuggability, with insightful logging around retries and errors in particular.
- An Opstrace instance.
- A decision: for which Opstrace tenant would you like to send data?
- An Opstrace tenant authentication token file (for the tenant of your choice). Also see concepts.
What follows is an example for a block matching all log entries, and for sending them to your Opstrace instance:
<match **>@type lokiurl <string>insecure_tls <boolean>bearer_token_file <filepath></match>
This is the tenant-specific data API endpoint of an Opstrace instance for ingesting metrics. The endpoint URL is constructed using the tenant name and the instance name:
Your Opstrace instance exposes the logs ingestion API securely. For authenticating the client towards the Opstrace instance., the client has to present a tenant-specific data API authentication token with every request.
To achieve that, we use the
bearer_token_file parameter. Example:
TLS is used for establishing a secure, private transport between the Fluentd instance and the Opstrace instance..
By default, Fluentd attempts to verify the certificate exposed by the Opstrace instance. against its trust store, for establishing authenticity of the Opstrace instance towards itself.
Depending on the
cert_issuer instance. configuration option (see configuration) you may want to disable server certificate verification for testing purposes, which can be achieved via
This is required for example when using
We recommend reading into the Fluentd Buffer Section documentation.
For example, when choosing a node-local Fluentd buffer of
@type file one can maximize the likelihood to recover from failures without losing valuable log data (the node-local persistent buffer can be flushed eventually -- Fluentd's default retrying timeout is 3 days). What follows is a configuration structure that serves as a starting point for more advanced flushing and buffering configuration:
<match **>@type lokiurl <string>insecure_tls <boolean>bearer_token_file <filepath><buffer>@type filepath <filepath>chunk_limit_size 512kbflush_interval 3s</buffer></match>
Note that Fluentd can also back up log entries that it failed to flush.