Add option to use host time server configuration for NTP check

All of our servers are configured to use a local time service to avoid making external calls for NTP. On premise, local servers are acting as proxies to external time sources. On cloud platforms, we use the time source provided by the hosting cloud providers.

Also, I have noticed that NTP queries from GCP to Datadog NTP servers time out several times a day.

It would be great if the NTP check used the time service configured on the host system by default. If NTP isn’t enabled/configured at the host level, it could then default to the datadog provided pool.

Notes:

Debian and Container-Optimized OS GCE images provided by Google use the local NTP service at metadata.google.internal (169.254.169.254) by default. https://cloud.google.com/compute/docs/instances/managing-instances#configure_ntp_for_your_instances

AWS has a similar service: 169.254.169.123 https://aws.amazon.com/blogs/aws/keeping-time-with-amazon-time-sync-service/

Other cloud providers certainly have too.

Bonus points: use those by default when running on cloud platforms.

[olivielpeau]

Implemented through https://github.com/DataDog/datadog-agent/pull/3979, which was just released in Agent v6.14.0. Closing.

We are not considering porting this enhancement to Agent v5 (see https://github.com/DataDog/datadog-agent/pull/3979#issuecomment-531793857)

[masci]

Thanks for the quick response, I’m coordinating with the Agent team to deliver this feature.