verminity opened a new issue #8616: URL: https://github.com/apache/pulsar/issues/8616
#### Expected behavior According to the [Prometheus text format documentation](https://prometheus.io/docs/instrumenting/exposition_formats/) > If the token is TYPE, exactly two more tokens are expected. The first is the metric name, and the second is either counter, gauge, histogram, summary, or untyped, defining the type for the metric of that name. **Only one TYPE line may exist for a given metric name.** It is expected that each `# TYPE` line returned by /metrics is unique #### Actual behavior When a cluster has more than one tenant some pulsar_ml metrics types appear more than once on /metrics The issue does not exist on Pulsar 2.5.1 and appears to have been introduced with "Expose managedLedgerCache, managedLedger, loadBalance metrics to Prometheus #6705 " in 2.6.0 #### Steps to reproduce [Start a standalone Pulsar in Docker](https://pulsar.apache.org/docs/en/standalone-docker/#start-pulsar-in-docker) Create a tenant Create a namespace below that tenant Create a persistent topic on that namespace GET http://localhost:8080/metrics Search the response for `# TYPE pulsar_ml_AddEntryBytesRate gauge` and you'll find two results (there are a number of duplicated types) The impact of this is trying to use Telegraf to read this metrics causes the following error: `Error in plugin: error reading metrics for http://localhost:8080/metrics: reading text format failed: second TYPE line for metric name "pulsar_ml_AddEntryBytesRate", or TYPE reported after samples` and most metrics are dropped #### System configuration **Pulsar version**: 2.6.2 ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
