22 Feb 2023
04:20 AM
- last edited on
22 Feb 2023
08:39 AM
by
MaciejNeumann
I have Dynatrace Managed server on-prem and also monitoring some Kubernetes on the cloud. Does Kubernetes pull the OneAgent image from Dynatrace Managed server each time an application pod is created, or is the OneAgent image cached within the Dynatrace operator?
Solved! Go to Solution.
22 Feb 2023 08:25 AM
As I know - each time. Others will correct me if I am wrong.
To reduce traffic you can upload image to Cloud repository and define image in DynaKube.yaml, but update version in that case will be manual.
22 Feb 2023 11:31 AM
IIRC the imagePullPolicy is set to always, so it will pull the image each time a pod is started, even if the image is already present.
23 Feb 2023 12:55 AM
That's good information!
Well, there is a concern for the bandwidth utilization as the image will be pulled each time a pod is started. What's the approximate file size of the image?
23 Feb 2023 08:00 AM
Hi @mengsuan_koe1 ,
which deployment option are you using in your Dynakube (classic full-stack vs cloud-native full-stack; assuming you are referring to full-stack monitoring)?
Ad "each time an application pod is created":
Are you referring to the OneAgent pod (default-namespace: Dynatrace, DaemonSet) or your applications pods (in your namespaces)?
For classic full-stack, the OneAgent image is downloaded by the container runtime (namespace: dynatrace). The injection into your application pods (your namespace) does not require any additional images/bandwidth.
For cloud-native, there is a difference between the OneAgent image and the code-modules image. While the OneAgent image is downloaded by the container runtime, the code-modules image is downloaded and managed by the CSI driver. The injection of the code-modules into your application pods is again leveraging the CSI driver (CSI volume) -> no additional image downloads/bandwidth. (The init container uses the container image from the Operator which is cached by the CR).
All container images are cached by the ActiveGate on your Dynatrace managed cluster.
Hope that helps!
17 Jan 2025 03:53 PM
Hello
I'm having a node disk pressure issue when a new pod is created each time and it is pulling the operator image all the time because its using imagePullPolicy is set to always and filing the disk space.
I'm using cloudNative with CSI. In the CSI yaml we have 4 containers all of them have imagePullPolicy is set to always.
is it okay to set imagePullPolicy is set to IfNotPresent in the yaml? since it always pull whataever the operator version is using.