FabledSteward/tests/plugins/host_agent/test_expand_metrics.py

"""Unit tests for _expand_sample_to_metrics — sample dict → PluginMetric rows.

No DB: PluginMetric objects are constructed in-memory, so we assert the
(metric_name, resource_name) → value contract directly.
"""
from datetime import datetime, timezone

from plugins.host_agent.routes import _expand_sample_to_metrics


def _index(rows):
    return {(r.metric_name, r.resource_name): r.value for r in rows}


def test_expand_emits_all_new_metric_families():
    sample = {
        "ts": "2026-06-16T00:00:00+00:00",
        "cpu_pct": 12.5,
        "cpu_cores": [10.0, 15.0],
        "mem": {"total_bytes": 100, "available_bytes": 40, "swap_used_bytes": 5,
                "cached_bytes": 20, "buffers_bytes": 3},
        "load": {"1m": 0.1, "5m": 0.2, "15m": 0.3},
        "uptime_secs": 1000,
        "storage": [{"mount": "/", "total_bytes": 100, "used_bytes": 50}],
        "net": [{"iface": "eth0", "rx_bps": 1000.0, "tx_bps": 500.0}],
        "diskio": [{"device": "sda", "read_bps": 2000.0, "write_bps": 1000.0}],
        "temps": [{"label": "Package", "celsius": 45.0}, {"label": "Core0", "celsius": 50.0}],
        "psi": {"mem_some_avg10": 1.5, "cpu_some_avg10": 0.0},
    }
    idx = _index(_expand_sample_to_metrics(sample, "alpha", datetime.now(timezone.utc)))

    # per-core CPU as sub-resources
    assert idx[("cpu_pct", "alpha:core0")] == 10.0
    assert idx[("cpu_pct", "alpha:core1")] == 15.0
    # richer memory
    assert idx[("mem_cached_bytes", "alpha")] == 20
    assert idx[("mem_buffers_bytes", "alpha")] == 3
    # network: per-iface + host total
    assert idx[("net_rx_bps", "alpha:net:eth0")] == 1000.0
    assert idx[("net_rx_bps", "alpha")] == 1000.0
    # disk I/O: per-device + host total
    assert idx[("disk_read_bps", "alpha:diskio:sda")] == 2000.0
    assert idx[("disk_write_bps", "alpha")] == 1000.0
    # temps: per-sensor + host max
    assert idx[("temp_c", "alpha:temp:Package")] == 45.0
    assert idx[("temp_c_max", "alpha")] == 50.0
    # PSI
    assert idx[("psi_mem_some_avg10", "alpha")] == 1.5


def test_expand_is_backcompat_with_old_agent_sample():
    """An old agent (no new keys) must still expand cleanly to legacy metrics."""
    sample = {"ts": "x", "cpu_pct": 5.0, "mem": {"total_bytes": 10, "available_bytes": 5}}
    rows = _expand_sample_to_metrics(sample, "h", datetime.now(timezone.utc))
    names = {r.metric_name for r in rows}
    assert "cpu_pct" in names
    assert not any(
        n.startswith(("net_", "disk_read", "disk_write", "psi_", "temp_", "mem_cached"))
        for n in names
    )


def test_every_resource_is_the_host_or_a_host_colon_part():
    """A host rename carries "<host>" and "<host>:…" (steward/core/renames.py
    HOST_SOURCES); a resource named any other way would stay behind."""
    from steward.core.renames import HOST_SOURCES
    sample = {
        "ts": "x", "cpu_pct": 1.0, "cpu_cores": [1.0],
        "mem": {"total_bytes": 10, "available_bytes": 5, "cached_bytes": 1, "buffers_bytes": 1},
        "load": {"1m": 0.1}, "uptime_secs": 1,
        "storage": [{"mount": "/", "total_bytes": 10, "used_bytes": 5}],
        "net": [{"iface": "eth0", "rx_bps": 1.0, "tx_bps": 1.0}],
        "diskio": [{"device": "sda", "read_bps": 1.0, "write_bps": 1.0}],
        "temps": [{"label": "Package", "celsius": 40.0}],
        "psi": {"cpu_some_avg10": 0.0},
    }
    sep = dict(HOST_SOURCES)["host_agent"]
    resources = {r.resource_name for r in _expand_sample_to_metrics(sample, "web01", datetime.now(timezone.utc))}
    assert len(resources) > 5
    assert {r for r in resources if r != "web01" and not r.startswith("web01" + sep)} == set()


def test_a_sensor_with_nothing_wired_to_it_does_not_set_the_host_s_temperature():
    """#5759: VDNT-Util01's AUXTIN0 read 110 °C flat while its CPU package
    moved 49-87, so the host read critical all week. Every sensor keeps its
    own row; only readings that can be real set the max."""
    sample = {"ts": "2026-10-14T21:00:00+00:00",
              "temps": [{"label": "SYSTIN", "celsius": 114.0},
                        {"label": "AUXTIN0", "celsius": 110.0},
                        {"label": "PCH_CHIP_CPU_MAX_TEMP", "celsius": 0.0},
                        {"label": "AUXTIN3", "celsius": -1.0},
                        {"label": "Package id 0", "celsius": 60.0}]}
    idx = _index(_expand_sample_to_metrics(sample, "util", datetime.now(timezone.utc)))
    assert idx[("temp_c_max", "util")] == 60.0
    assert idx[("temp_c", "util:temp:AUXTIN0")] == 110.0
    assert idx[("temp_c", "util:temp:PCH_CHIP_CPU_MAX_TEMP")] == 0.0


def test_stalled_mounts_are_counted_and_have_no_figures():
    """#5760: a mount the agent couldn't read in time."""
    sample = {"ts": "2026-10-14T21:00:00+00:00",
              "storage": [{"mount": "/", "total_bytes": 100, "used_bytes": 50},
                          {"mount": "/nfs/media", "stalled": True}]}
    idx = _index(_expand_sample_to_metrics(sample, "d3", datetime.now(timezone.utc)))
    assert idx[("mounts_stalled", "d3")] == 1
    assert idx[("disk_used_pct_worst", "d3")] == 50.0
    assert ("disk_used_pct", "d3:/nfs/media") not in idx


def test_a_wired_board_input_still_counts():
    """#5759: below 100 °C a board input is a real reading."""
    from plugins.host_agent.routes import counts_as_hottest
    assert counts_as_hottest("SYSTIN", 38.0) and counts_as_hottest("AUXTIN1", 60.0)
    assert not counts_as_hottest("SYSTIN", 114.0) and not counts_as_hottest("PCH_CHIP_TEMP", 0.0)
    assert counts_as_hottest("Package id 0", 101.0)     # a CPU past 100 is a hot CPU


def test_the_agent_s_own_footprint_is_recorded():
    """#5764: a runaway agent shows on the record before it goes silent."""
    sample = {"ts": "2026-10-14T22:00:00+00:00",
              "agent": {"rss_bytes": 18_000_000, "cpu_pct": 1.5}}
    idx = _index(_expand_sample_to_metrics(sample, "d3", datetime.now(timezone.utc)))
    assert idx[("agent_rss_bytes", "d3")] == 18_000_000
    assert idx[("agent_cpu_pct", "d3")] == 1.5