You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Related to #17600, I see that some of the metric names for the cockroachdb integration are arguably wrong. For example, cockroachdb includes a metric called admission.wait_queue_length.kv. This name makes sense, because admission is a cockroachdb subsystem, and wait_queue_length is a property of that system, and kv is a subcategory of wait_queue_length. However, the Datadog integration renames this metric to admission.wait.queue.length.kv. This doesn't make sense to me: wait, queue, and length aren't nested concepts, they're a single concept called wait_queue_length. More to the point, the cockroachdb authors gave that metric its name for a reason, and I would think that Datadog should generally respect upstream naming conventions.
I would be happy to send a PR to fix the cockroachdb metric names, but I think the real solution is to check in the code that generates the metric names, so that we can consistently update metrics with new cockroachdb releases. Then if metric names need to be changed, we can fix them at the level of the automation code and not review individual metric names by hand.
Thank you! We'll also send a support ticket so we have something that follows the official support process. In my opinion Datadog should respect the metric names that the application emits since that's what the application documents and what operators and developers expect.
It's fine for Datadog to add a prefix to the metrics that links the metric to the integration and prevent naming collisions with other integrations but that's the only mutation that should occur.
I'm proposing that Datadog update this integration to emit metrics in the format PREFIX.METRIC_NAME where PREFIX is the integration name (i.e, cockroachdb) and METRIC_NAME is the metric name as emitted from the application (i.e., admission.wait_queue_length.kv).
Related to #17600, I see that some of the metric names for the cockroachdb integration are arguably wrong. For example, cockroachdb includes a metric called admission.wait_queue_length.kv. This name makes sense, because
admission
is a cockroachdb subsystem, andwait_queue_length
is a property of that system, andkv
is a subcategory ofwait_queue_length
. However, the Datadog integration renames this metric to admission.wait.queue.length.kv. This doesn't make sense to me:wait
,queue
, andlength
aren't nested concepts, they're a single concept calledwait_queue_length
. More to the point, the cockroachdb authors gave that metric its name for a reason, and I would think that Datadog should generally respect upstream naming conventions.I would be happy to send a PR to fix the cockroachdb metric names, but I think the real solution is to check in the code that generates the metric names, so that we can consistently update metrics with new cockroachdb releases. Then if metric names need to be changed, we can fix them at the level of the automation code and not review individual metric names by hand.
@FlorentClarret @sudomateo
The text was updated successfully, but these errors were encountered: