domino_admin_toolkit.analyzers.node_coverage module

Shared coverage verdict for node-level checks backed by node-exporter.

A node-exporter-backed frame only covers nodes where the exporter runs and is scraped, so without this verdict a check cannot tell “every node is healthy” from “the nodes I could see are healthy” — on one production cluster that meant judging 9 of 107 nodes. The analyzer compares the check’s own frame against the kube-state-metrics inventory (domino_admin_toolkit.lib.node_pool.fetch_node_inventory()) and emits one result.

pydantic model domino_admin_toolkit.analyzers.node_coverage.NodeCoverageAnalyzer

Bases: DataFrameAnalyzerBase

Emits one result describing how much of the cluster a node-level DataFrame observed.

PASS at full coverage, WARN on a partial gap, ERROR once the gap is large enough to invalidate the check’s own verdict, and WARN when the inventory itself is unavailable.

Fields:

• inventory (tuple[str, ...])

• min_coverage_ratio (float)

• min_missing_for_error (int)

• source (str)

field inventory: tuple[str, ...] = ()

field min_coverage_ratio: float = 0.9

field min_missing_for_error: int = 2

field source: str = 'node-exporter'

analyze(data)

Compare the node column of data against the inventory and record one verdict.

Return type:

list[CheckResult]

name: ClassVar[str] = 'NodeCoverageAnalyzer'

domino_admin_toolkit.analyzers.node_coverage.add_node_coverage_result(runner, data, inventory, *, min_coverage_ratio=0.9, source='node-exporter')

Append the coverage verdict to the same Runner that holds the check’s own results.

Constructed directly rather than via runner.check_df(), which instantiates analyzers with no arguments and so cannot pass the inventory in. Results accumulate on the Runner, so a coverage ERROR fails the check through present_and_assert().

Return type:

None