error in grpc streaming ingest_measurements

[ezychlagre]

Describe the bug
On a ovh k8s cluster, I deployed an ALUMet client on each cluster's node (2) and 1 ALUMet server. Sometimes, few second after the deployment (using an helm chart), all ALUMet's client crashed with the following error:
ERROR alumet::pipeline::elements::output] Error when asynchronously writing to relay-client/output/grpc-measurements (will stop running): error in grpc streaming ingest_measurements: status status: Cancelled, message: "h2 protocol error: http2 error", details: [], metadata: MetadataMap { headers: {} }

Environment information
OS: k8s on ovh
Alumet version²: 0.6.1 + energy-estimation-tdp plugin
Enabled plugins:

• EnergyEstimationTdpPlugin v0.1.0
  - k8s v0.1.0
  - procfs v0.1.0
  - relay-client v0.4.0
  - socket-control v0.2.0

²: please include the commit hash if it's not a release
2611b11

To Reproduce
Steps to reproduce the behavior:

1. deploy the ALUMet helm chart but the error is not reproductable every time ! I reproduced at least 3 times on ovh k8s cluster.

I did several deploiement tests this afternoon on ovh ck8s luster and the results are:

1. 16h03: reproduce the error,
2. 16h18: no error
3. 16h27: reproduce the error,
4. 16h31: no error
5. 14h41: no error
6. 16h57: no error
7. 17h05: no error

Expected behavior
the ALUMet pipepline crashed, we get a grpc error

Logs / Output
Attached logs file of server and client part.
alumet-client-ovh-demo-ns-demo-test-2024-11-13.txt
alumet-server-ovh-demo-ns-demo-test-2024-11-13.txt
Below a screenshot influxdb dashboard but not at the same times as the above log files