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
Datadog's helm deployment for k8s makes it difficult to customize which checks I use for which hosts, so I need a tier wide check. My cluster is mixed CPU and GPU hosts. If I deploy this integration, self.event is triggered each check for all my CPU enabled hosts. I would like configuration to disable sending self.event messages for the plugin. This will reduce the amount of noise I see in datadog error events.
The text was updated successfully, but these errors were encountered:
Hi,
Datadog's helm deployment for k8s makes it difficult to customize which checks I use for which hosts, so I need a tier wide check. My cluster is mixed CPU and GPU hosts. If I deploy this integration,
self.event
is triggered each check for all my CPU enabled hosts. I would like configuration to disable sendingself.event
messages for the plugin. This will reduce the amount of noise I see in datadog error events.The text was updated successfully, but these errors were encountered: