Skip to content

Post-mission DumpData/GetRawMetrics crashes when cluster does not resolve .local TLD #399

Description

@tomerweller

Observed behavior

When running Supercluster missions on a Namespace k3s cluster, the mission logic completes successfully but the teardown metrics dump crashes with:

System.Net.WebException: Name or service not known (ssc-xxxxx-yyyyyy.local:80)

This happens during DumpPeerMetrics → GetRawMetrics → Peer.fetch(), which resolves the per-pod ingress hostname (e.g. ssc-1840z-d05de1.local) to scrape admin HTTP metrics before tearing down Kubernetes resources. The .local TLD does not resolve in non-standard DNS environments like Namespace k3s clusters.

Impact

Missions appear to fail (SSC exits non-zero) even when the actual consensus/loadgen logic succeeded. This prevents automated pass/fail gating for mixed-image or custom-image missions on non-SDF Kubernetes clusters.

Suggested fix

Use the per-pod Kubernetes service DNS name instead of the ingress .local hostname for the post-mission metrics scrape. The svc.cluster.local names already work and are used elsewhere in the SSC config generation.

Alternatively, make the ingress domain suffix configurable (currently hardcoded to .local via StellarCoreCfg.fs / StellarKubeSpecs.fs).

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions