bvandeusenandClaude Opus 5.5
b29b160baa
- Overview: a verdict line when something is wrong ('1 alert firing, 1
monitor down.'); its picker description now says what it lists (alerts,
monitors down, anomalies). Host figures have their home in Hosts.
Comments no longer name the Status page that was folded into Monitors.
- Monitors widget: a monitor not checked yet says when it was added.
- Host page: filesystems fullest first, by the worse of space and inodes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
203 lines
11 KiB
Python
203 lines
11 KiB
Python
"""The dashboard starts from a curated layout, the duplicate widgets are gone,
|
|
and the Overview says what each number counts (#5213)."""
|
|
import re
|
|
from datetime import datetime, timedelta, timezone
|
|
from types import SimpleNamespace
|
|
|
|
from quart import render_template, session
|
|
|
|
from steward.core.status import StatusEntry
|
|
from steward.core.widgets import RECOMMENDED, RETIRED, WIDGET_REGISTRY, recommended_layout
|
|
from steward.status.routes import overview_alerts
|
|
from tests.test_style_tokens import hex_colours
|
|
|
|
ALL_ON = {p: {"enabled": True} for p in ("docker", "host_agent", "traefik", "unifi", "snmp")}
|
|
|
|
|
|
def _overlap(a, b):
|
|
return (a["x"] < b["x"] + b["w"] and b["x"] < a["x"] + a["w"]
|
|
and a["y"] < b["y"] + b["h"] and b["y"] < a["y"] + a["h"])
|
|
|
|
|
|
def test_retired_widgets_point_at_the_one_that_replaced_them():
|
|
for old, new in RETIRED.items():
|
|
assert old not in WIDGET_REGISTRY and new in WIDGET_REGISTRY
|
|
assert all(key in WIDGET_REGISTRY for key, *_ in RECOMMENDED)
|
|
|
|
|
|
def test_the_recommended_layout_fills_rows_without_overlap():
|
|
items = recommended_layout(ALL_ON)
|
|
assert [i["key"] for i in items] == [key for key, *_ in RECOMMENDED]
|
|
# The Overview and the Alerts side by side first (#5510), then the fleet.
|
|
assert items[0] == {"key": "status_overview", "x": 0, "y": 0, "w": 6, "h": 4, "config": {}}
|
|
assert items[1] == {"key": "alerts", "x": 6, "y": 0, "w": 6, "h": 4, "config": {}}
|
|
assert [(i["key"], i["x"], i["y"]) for i in items[2:4]] == [
|
|
("hosts_overview", 0, 4), ("monitors", 6, 4)]
|
|
for i, a in enumerate(items):
|
|
assert a["x"] + a["w"] <= 12
|
|
assert not any(_overlap(a, b) for b in items[i + 1:]), a
|
|
|
|
|
|
def test_the_recommended_layout_skips_what_is_not_enabled():
|
|
items = recommended_layout({})
|
|
assert [i["key"] for i in items] == ["status_overview", "alerts", "hosts_overview", "monitors"]
|
|
assert [(i["x"], i["y"]) for i in items] == [(0, 0), (6, 0), (0, 4), (6, 4)]
|
|
|
|
|
|
def test_overview_alerts_put_the_breaching_first():
|
|
from steward.models.alerts import AlertStateEnum
|
|
ack = SimpleNamespace(state=AlertStateEnum.acknowledged, last_transition_at=None, no_data_since=None)
|
|
silent = SimpleNamespace(state=AlertStateEnum.inactive, last_transition_at=None,
|
|
no_data_since=datetime(2026, 10, 7, tzinfo=timezone.utc))
|
|
firing = SimpleNamespace(state=AlertStateEnum.firing, last_transition_at=datetime(2026, 10, 6, tzinfo=timezone.utc),
|
|
no_data_since=None)
|
|
rows = overview_alerts([(ack, "r-ack"), (silent, "r-silent"), (firing, "r-firing")])
|
|
assert [(r["rule"], r["label"]) for r in rows] == [
|
|
("r-firing", "firing"), ("r-silent", "no_data"), ("r-ack", "acknowledged")]
|
|
assert rows[1]["since"] == silent.no_data_since # silent since the series stopped
|
|
assert rows[0]["since"] == firing.last_transition_at
|
|
|
|
|
|
async def test_the_overview_labels_its_counts_and_lists_what_is_wrong(app):
|
|
now = datetime.now(timezone.utc)
|
|
rule = SimpleNamespace(id="r1", name="nas disk full")
|
|
alerts = [{"rule": rule, "label": "firing", "since": now - timedelta(minutes=12)}]
|
|
entries = [
|
|
StatusEntry(kind="http", key="1", name="web", status="up", detail_url="/monitors/1"),
|
|
StatusEntry(kind="tcp", key="2", name="ssh-nas", status="down", detail_url="/monitors/2",
|
|
since=now - timedelta(hours=2)),
|
|
]
|
|
async with app.test_request_context("/status/widget"):
|
|
session["user_id"] = "u1"
|
|
page = await render_template("status/widget.html", entries=entries, up=1, down=1, pending=0,
|
|
total_hosts=3, alerts=alerts, widget_id="1")
|
|
assert ">Monitors<" in page and ">Alerts<" in page and ">firing<" in page and ">hosts<" in page
|
|
assert '<span class="badge badge-crit">Firing</span>' in page
|
|
assert 'href="/alerts/history?rule=r1"' in page and "12m ago" in page
|
|
assert '<span class="badge badge-crit">Down</span>' in page and "2h ago" in page
|
|
assert page.index("nas disk full") < page.index("ssh-nas") # alerts, then monitors
|
|
assert ">pending<" not in page # a zero pending count is not shown
|
|
# #5810: a verdict says what is wrong before the list says where.
|
|
assert '<div class="w-verdict tone-crit">1 alert firing, 1 monitor down.</div>' in page
|
|
assert page.index("1 alert firing, 1 monitor down.") < page.index("nas disk full")
|
|
assert not hex_colours(page), hex_colours(page)
|
|
|
|
|
|
async def test_the_overviews_overflow_links_to_alerts_and_to_monitors(app):
|
|
"""#5510: what doesn't fit goes to its own list; /status/ is gone."""
|
|
now = datetime.now(timezone.utc)
|
|
alerts = [{"rule": SimpleNamespace(id=f"r{i}", name=f"rule {i}"), "label": "firing", "since": now}
|
|
for i in range(7)]
|
|
entries = [StatusEntry(kind="tcp", key=str(i), name=f"m{i}", status="down", detail_url=f"/monitors/{i}")
|
|
for i in range(10)]
|
|
async with app.test_request_context("/status/widget"):
|
|
session["user_id"] = "u1"
|
|
page = await render_template("status/widget.html", entries=entries, up=0, down=10, pending=0,
|
|
total_hosts=1, alerts=alerts, widget_id="1")
|
|
assert '<a href="/alerts/">+1 more alerts</a> · <a href="/monitors/">+2 more monitors</a>' in page
|
|
assert "/status/" not in page
|
|
async with app.test_request_context("/status/widget"):
|
|
page = await render_template("status/widget.html", entries=entries[:3], up=0, down=3, pending=0,
|
|
total_hosts=1, alerts=alerts, widget_id="1")
|
|
assert "+1 more alerts" in page and "more monitors" not in page
|
|
assert 'href="/alerts/"' not in page # the share view has no links
|
|
|
|
|
|
async def test_a_quiet_overview_says_so(app):
|
|
async with app.test_request_context("/status/widget"):
|
|
page = await render_template("status/widget.html", entries=[], up=0, down=0, pending=0,
|
|
total_hosts=1, alerts=[], widget_id="1")
|
|
assert "No monitors configured, and no alert is firing." in page
|
|
assert ">host<" in page and 'href="/alerts/"' not in page # the share view has no links
|
|
|
|
|
|
def _host(hid, name):
|
|
return SimpleNamespace(id=hid, name=name)
|
|
|
|
|
|
def test_hosts_sort_worst_first():
|
|
"""#5251: a down monitor, then a reading past critical, then a quiet agent,
|
|
then a warning, then the healthy, by name; a host with nothing last."""
|
|
from steward.core.settings import to_thresholds_cfg
|
|
from steward.hosts.service import host_glance
|
|
th = to_thresholds_cfg({})
|
|
hosts = [_host("a", "alpha"), _host("b", "busy"), _host("d", "dead"), _host("q", "quiet"),
|
|
_host("w", "warm"), _host("n", "nothing")]
|
|
went = datetime(2026, 10, 15, 9, 0, tzinfo=timezone.utc)
|
|
status = {"d": {"state": "down", "down_since": went}, "a": {"state": "up", "down_since": None}}
|
|
agent = {"alpha": {"cpu_pct": 10.0, "fresh": True}, "busy": {"cpu_pct": 99.0, "fresh": True},
|
|
"quiet": {"cpu_pct": 99.0, "fresh": False},
|
|
"warm": {"cpu_pct": 10.0, "load_per_core": 85, "fresh": True}}
|
|
rows = host_glance(hosts, status, agent, th)
|
|
assert [(r["host"].name, r["state"]) for r in rows] == [
|
|
("dead", "down"), ("busy", "crit"), ("quiet", "stale"), ("warm", "warn"),
|
|
("alpha", "up"), ("nothing", "unknown")]
|
|
assert rows[1]["worst"][0] == "CPU" and rows[3]["worst"][0] == "Load"
|
|
assert rows[2]["worst"] is None # a quiet agent's old figures aren't graded
|
|
assert rows[0]["down_since"] == went and rows[4]["down_since"] is None # #5810
|
|
|
|
|
|
async def test_a_down_host_says_for_how_long(app):
|
|
"""#5810: the problem row carries its since, as the Hosts list does."""
|
|
rows = [{"host": _host("d", "dead"), "state": "down", "worst": None, "agent": {},
|
|
"down_since": datetime.now(timezone.utc) - timedelta(hours=3)}]
|
|
async with app.test_request_context("/hosts/overview/widget"):
|
|
page = await render_template("hosts/overview_widget.html", rows=rows)
|
|
page = " ".join(page.split())
|
|
assert "a monitor is down for" in page and ">3h<" in page and " ago" not in page
|
|
|
|
|
|
async def test_the_hosts_widget_gives_a_verdict_and_load_against_cores(app):
|
|
from datetime import datetime, timezone
|
|
rows = [
|
|
{"host": _host("h2", "VDNT-Util01"), "state": "crit", "worst": ("CPU", 99.0, "crit"),
|
|
"agent": {"cpu_pct": 99.0, "mem_used_pct": 28.0, "disk_root": 52.0, "load_1m": 16.77,
|
|
"cores": 8, "load_per_core": 210, "fresh": True}},
|
|
{"host": _host("h1", "VDNT-Docker02"), "state": "up", "worst": None,
|
|
"agent": {"cpu_pct": 1.0, "mem_used_pct": 4.0, "disk_root": 44.0, "load_1m": 0.92,
|
|
"cores": 4, "load_per_core": 23, "fresh": True}},
|
|
{"host": _host("h3", "nas"), "state": "stale", "worst": None,
|
|
"agent": {"cpu_pct": 5.0, "fresh": False, "last_at": datetime(2026, 10, 7, tzinfo=timezone.utc)}},
|
|
]
|
|
async with app.test_request_context("/hosts/overview/widget"):
|
|
session["user_id"] = "u1"
|
|
page = await render_template("hosts/overview_widget.html", rows=rows)
|
|
assert "2 of 3 hosts need a look." in page
|
|
assert page.index("VDNT-Util01") < page.index("VDNT-Docker02")
|
|
assert "16.8 / 8 cores" in page and "0.9 / 4 cores" in page
|
|
assert "agent last reported" in page
|
|
assert len(re.findall(r"<th[ >]", page)) == 5 and ">ping<" not in page and "24h" not in page
|
|
assert not hex_colours(page), hex_colours(page)
|
|
|
|
|
|
async def test_one_busy_host_is_named_in_the_verdict(app):
|
|
rows = [{"host": _host("h2", "web-1"), "state": "crit", "worst": ("CPU", 99.0, "crit"),
|
|
"agent": {"cpu_pct": 99.0, "fresh": True}}]
|
|
async with app.test_request_context("/hosts/overview/widget"):
|
|
page = await render_template("hosts/overview_widget.html", rows=rows)
|
|
assert "web-1:" in page and "CPU high." in page and "tone-crit" in page
|
|
|
|
|
|
async def test_a_widget_polls_at_its_own_pace(app):
|
|
defn = {"poll": True, "poll_seconds": 300, "detail_url": "/x"}
|
|
item = SimpleNamespace(db=SimpleNamespace(id=7, grid_x=0, grid_y=0, grid_w=6, grid_h=5),
|
|
title="Slow", defn=defn, hx_url="/slow/widget")
|
|
async with app.test_request_context("/"):
|
|
page = await render_template("dashboard/_grid_item.html", item=item, can_edit=False, poll_interval=30)
|
|
assert "every 300s" in page
|
|
defn.pop("poll_seconds")
|
|
async with app.test_request_context("/"):
|
|
page = await render_template("dashboard/_grid_item.html", item=item, can_edit=False, poll_interval=30)
|
|
assert "every 30s" in page
|
|
|
|
|
|
async def test_a_grid_item_carries_its_saved_layout(app):
|
|
"""#5249: the saved 12-column place and height, which the page fits to
|
|
content in view mode and restores in edit mode."""
|
|
item = SimpleNamespace(db=SimpleNamespace(id=7, grid_x=6, grid_y=3, grid_w=6, grid_h=5),
|
|
title="Uptime", defn={"poll": False, "detail_url": "/x"}, hx_url="/uptime/widget")
|
|
async with app.test_request_context("/"):
|
|
page = await render_template("dashboard/_grid_item.html", item=item, can_edit=False, poll_interval=30)
|
|
assert 'data-x="6" data-y="3"' in page and 'data-w="6" data-h="5"' in page
|
|
assert 'gs-x="6" gs-y="3"' in page
|