def test_auto_refids_preserves_provided_ids():
"""
auto_ref_ids() provides refIds for all targets without refIds already
set.
"""
dashboard = G.Dashboard(
title="Test dashboard",
rows=[
G.Row(panels=[
G.Graph(
title="CPU Usage by Namespace (rate[5m])",
targets=[
G.Target(
expr='whatever #Q',
legendFormat='{{namespace}}',
),
G.Target(
expr='hidden whatever',
legendFormat='{{namespace}}',
refId='Q',
),
G.Target(expr='another target'),
],
).auto_ref_ids()
]),
],
)
assert dashboard.rows[0].panels[0].targets[0].refId == 'A'
assert dashboard.rows[0].panels[0].targets[1].refId == 'Q'
assert dashboard.rows[0].panels[0].targets[2].refId == 'B'
dashboard = G.Dashboard(
title="Test dashboard",
panels=[
G.Graph(title="CPU Usage by Namespace (rate[5m])",
dataSource="My data source",
targets=[
G.Target(
expr='whatever #Q',
legendFormat='{{namespace}}',
),
G.Target(
expr='hidden whatever',
legendFormat='{{namespace}}',
refId='Q',
),
G.Target(expr='another target'),
],
yAxes=G.YAxes(
G.YAxis(format=G.SHORT_FORMAT, label="CPU seconds"),
G.YAxis(format=G.SHORT_FORMAT),
),
gridPos=G.GridPos(h=1, w=24, x=0, y=8)).auto_ref_ids()
],
).auto_panel_ids()
assert dashboard.panels[0].targets[0].refId == 'A'
assert dashboard.panels[0].targets[1].refId == 'Q'
assert dashboard.panels[0].targets[2].refId == 'B'
def PromGraph(data_source, title, expressions, **kwargs):
"""Create a graph that renders Prometheus data.
:param str data_source: The name of the data source that provides
Prometheus data.
:param title: The title of the graph.
:param expressions: List of tuples of (legend, expr), where 'expr' is a
Prometheus expression. Or a list of dict where keys are Target's args.
:param kwargs: Passed on to Graph.
"""
letters = string.ascii_uppercase
expressions = list(expressions)
if len(expressions) > len(letters):
raise ValueError(
'Too many expressions. Can support at most {}, but got {}'.format(
len(letters), len(expressions)))
if all(isinstance(expr, dict) for expr in expressions):
targets = [
G.Target(refId=refId, **args)
for (args, refId) in zip(expressions, letters)
]
else:
targets = [
G.Target(expr, legend, refId=refId)
for ((legend, expr), refId) in zip(expressions, letters)
]
return G.Graph(title=title,
dataSource=data_source,
targets=targets,
**kwargs)
def filesystem_usage(datasource):
return G.Graph(
title="Filesystem Usage",
dataSource=datasource,
xAxis=X_TIME,
yAxes=[
G.YAxis(
format="percent",
),
G.YAxis(
show=False,
),
],
targets=[
G.Target(
# Get the proportion used of each filesystem on a volume from
# a PersistentVolumeClaim on each node of the cluster. It's
# hard to figure out the role each filesystem serves from this
# graph (since all we get is the PVC name). Better than
# nothing, though. Hopefully later we can do better.
expr="""
100
* filesystem_used_bytes{volume=~"pvc-.*"}
/ filesystem_size_bytes{volume=~"pvc-.*"}
""",
legendFormat="{{volume}}",
refId="A",
),
],
)
def process_open_fds(datasource):
return G.Graph(
title="Open File Descriptors",
dataSource=datasource,
xAxis=X_TIME,
yAxes=[
G.YAxis(
format="none",
label="Count",
),
G.YAxis(
show=False,
),
],
targets=[
G.Target(
expr="""
process_open_fds{pod=~".+"}
""",
refId="A",
legendFormat="{{pod}}",
),
],
)
def memory_usage(datasource):
return G.Graph(
title="Memory Usage",
dataSource=datasource,
xAxis=X_TIME,
yAxes=[
G.YAxis(
# 2 ^ 30 bytes
format="gbytes",
label="Memory",
),
G.YAxis(
show=False,
),
],
targets=[
G.Target(
expr="""
sum(machine_memory_bytes) / 2 ^ 30
""",
legendFormat="Total Physical Memory",
refId="A",
),
G.Target(
expr="""
rss:container_memory:total / 2 ^ 30
""",
legendFormat="Total Container RSS",
refId="B",
),
],
)
def test_auto_id():
"""auto_panel_ids() provides IDs for all panels without IDs already set."""
dashboard = G.Dashboard(
title="Test dashboard",
rows=[
G.Row(panels=[
G.Graph(
title="CPU Usage by Namespace (rate[5m])",
dataSource="My data source",
targets=[
G.Target(
expr='whatever',
legendFormat='{{namespace}}',
refId='A',
),
],
yAxes=[
G.YAxis(format=G.SHORT_FORMAT, label="CPU seconds"),
G.YAxis(format=G.SHORT_FORMAT),
],
)
]),
],
).auto_panel_ids()
assert dashboard.rows[0].panels[0].id == 1
def test_auto_refids_preserves_provided_ids():
"""
auto_ref_ids() provides refIds for all targets without refIds already
set.
"""
dashboard = G.Dashboard(
title="Test dashboard",
rows=[
G.Row(panels=[
G.Graph(
title="CPU Usage by Namespace (rate[5m])",
targets=[
G.Target(
expr='whatever #Q',
legendFormat='{{namespace}}',
),
G.Target(
expr='hidden whatever',
legendFormat='{{namespace}}',
refId='Q',
),
G.Target(
expr='another target'
),
],
).auto_ref_ids()
]),
],
)
assert dashboard.rows[0].panels[0].targets[0].refId == 'A'
assert dashboard.rows[0].panels[0].targets[1].refId == 'Q'
assert dashboard.rows[0].panels[0].targets[2].refId == 'B'
def unhandled_errors(datasource):
return G.Graph(
title="Unhandled Errors",
dataSource=datasource,
xAxis=X_TIME,
yAxes=[
G.YAxis(
format="none",
label="Count",
),
G.YAxis(
show=False,
),
],
targets=[
G.Target(
expr="""
sum(s4_unhandled_error_counter)
""",
refId="A",
legendFormat="Total Unhandled Errors",
),
],
)
def last_convergence(datasource):
return G.Graph(
title="Since Last Convergence",
dataSource=datasource,
xAxis=X_TIME,
yAxes=[
G.YAxis(
format="none",
label="Period",
),
G.YAxis(
show=False,
),
],
targets=[
G.Target(
expr="""
time()
- max(
s4_last_convergence_succeeded{
pod=~"subscription-converger-.*"
}
)
""",
refId="A",
legendFormat="Time Since Last Convergence Success",
),
],
)
def test_graph_panel():
data_source = 'dummy data source'
targets = ['dummy_prom_query']
title = 'dummy title'
graph = G.Graph(data_source, targets, title)
data = graph.to_json_data()
assert data['targets'] == targets
assert data['datasource'] == data_source
assert data['title'] == title
assert 'alert' not in data
def simple_graph(title, exprs, yAxes=None):
if not isinstance(exprs, (list, tuple)):
exprs = [exprs]
return g.Graph(
title=title,
dataSource="$source",
# One graph per row.
span=g.TOTAL_SPAN,
targets=[g.Target(expr=expr) for expr in exprs],
yAxes=yAxes or g.YAxes(),
tooltip=DECREASING_ORDER_TOOLTIP,
)
def AddGraphPanel(self, title: Text, raw_sql: Text, y_axis_title: Text):
self.AddPanel(
core.Graph(
title=title,
targets=[
core.SqlTarget(
rawSql=raw_sql,
format=core.TABLE_TARGET_FORMAT,
),
],
yAxes=core.YAxes(core.YAxis(format=y_axis_title), ),
))
def test_graphite_target_full():
dashboard = G.Graph(title="Graphite target full test",
dataSource="graphite datasource",
targets=[
G.GraphiteTarget(refId="A", target="foo.bar"),
G.GraphiteTarget(refId="B",
target="sumSeries(#A,foo2.bar2)")
])
dashboard.resolve_graphite_targets()
for target in dashboard.targets:
assert target.targetFull != ""
assert not re.findall("#[A-Z]", target.targetFull)
def s4_customer_deployments(datasource):
return G.Graph(
title="Customer Deployments",
dataSource=datasource,
xAxis=X_TIME,
yAxes=[
G.YAxis(
format="none",
label="Total Customer Deployments",
min=0,
max=100,
),
G.YAxis(
show=False,
),
],
targets=[
G.Target(
# Each replicaset and pod end up with their own series. Label
# these more succinctly. Leave them distinct in case it is
# interesting to see where restarts have happened.
expr="""
label_replace(
s4_deployment_gauge{pod=~"subscription-converger-.*"},
"shortpod",
"# Deploys ($1)",
"pod",
"subscription-converger-(.*)"
)
""",
refId="A",
legendFormat="{{shortpod}}",
),
G.Target(
# As above.
expr="""
label_replace(
s4_running_pod_gauge{pod=~"subscription-converger-.*"},
"shortpod",
"# Running ($1)",
"pod",
"subscription-converger-(.*)"
)
""",
refId="B",
legendFormat="{{shortpod}}",
),
],
)
def run(self):
templateList = [
G.Template(default="",
dataSource="default",
name="serverid",
label="ServerID",
query="label_values(serverid)")
]
dashboard = G.Dashboard(title=self.options.title,
templating=G.Templating(list=templateList))
# Simple table processing - could be enhanced to use GridPos etc.
for metric in metrics:
if 'section' in metric:
dashboard.rows.append(
G.Row(title=metric['section'], showTitle=True))
continue
if 'row' in metric:
dashboard.rows.append(G.Row(title='', showTitle=False))
continue
graph = G.Graph(title=metric['title'],
dataSource='default',
maxDataPoints=1000,
legend=G.Legend(show=True,
alignAsTable=True,
min=True,
max=True,
avg=True,
current=True,
total=True,
sort='max',
sortDesc=True),
yAxes=G.single_y_axis())
ref_id = 'A'
for texp in metric['expr']:
graph.targets.append(G.Target(expr=texp, refId=ref_id))
ref_id = chr(ord(ref_id) + 1)
dashboard.rows[-1].panels.append(graph)
# Auto-number panels - returns new dashboard
dashboard = dashboard.auto_panel_ids()
s = io.StringIO()
write_dashboard(dashboard, s)
print("""{
"dashboard": %s
}
""" % s.getvalue())
def _row(title):
return core.Row(panels=[
core.Graph(title=title,
dataSource='prometheus',
targets=[
core.Target(
expr=title,
legendFormat='{{namespace}}',
),
],
yAxes=[
core.YAxis(format=core.NO_FORMAT),
core.YAxis(format=core.SHORT_FORMAT),
])
])
def test_graph_panel_threshold():
data_source = 'dummy data source'
targets = ['dummy_prom_query']
title = 'dummy title'
thresholds = [
G.GraphThreshold(20.0),
G.GraphThreshold(40.2, colorMode="ok")
]
graph = G.Graph(data_source, targets, title, thresholds=thresholds)
data = graph.to_json_data()
assert data['targets'] == targets
assert data['datasource'] == data_source
assert data['title'] == title
assert 'alert' not in data
assert data['thresholds'] == thresholds
def test_serialization_cloudwatch_metrics_target():
"""Serializing a graph doesn't explode."""
graph = G.Graph(
title="Lambda Duration",
dataSource="Cloudwatch data source",
targets=[
C.CloudwatchMetricsTarget(),
],
id=1,
yAxes=G.YAxes(
G.YAxis(format=G.SHORT_FORMAT, label="ms"),
G.YAxis(format=G.SHORT_FORMAT),
),
)
stream = StringIO()
_gen.write_dashboard(graph, stream)
assert stream.getvalue() != ''
def test_serialization_humio_metrics_target():
"""Serializing a graph doesn't explode."""
graph = G.Graph(
title="Humio Logs",
dataSource="Humio data source",
targets=[
H.HumioTarget(),
],
id=1,
yAxes=G.YAxes(
G.YAxis(format=G.SHORT_FORMAT, label="ms"),
G.YAxis(format=G.SHORT_FORMAT),
),
)
stream = StringIO()
_gen.write_dashboard(graph, stream)
assert stream.getvalue() != ''
def test_graph_panel_alert():
data_source = 'dummy data source'
targets = ['dummy_prom_query']
title = 'dummy title'
alert = [
G.AlertCondition(G.Target(), G.Evaluator('a', 'b'), G.TimeRange('5', '6'), 'd', 'e')
]
thresholds = [
G.GraphThreshold(20.0),
G.GraphThreshold(40.2, colorMode="ok")
]
graph = G.Graph(data_source, targets, title, thresholds=thresholds, alert=alert)
data = graph.to_json_data()
assert data['targets'] == targets
assert data['datasource'] == data_source
assert data['title'] == title
assert data['alert'] == alert
assert data['thresholds'] == []
def cpu_usage(datasource, intervals):
return G.Graph(
title="CPU usage",
dataSource=datasource,
xAxis=X_TIME,
yAxes=[
G.YAxis(
format="percent",
label="Average",
min=0,
max=100,
),
G.YAxis(
format="percent",
label="Average",
),
],
targets=list(
G.Target(
# CPU usage (as a percentage of maximum possible) averaged
# over a period is given as 100 times the sum (over all
# containers) of the rate of increase (in seconds) divided by
# the maximum possible increase (1 second per CPU).
#
# The sums are taken from recording rules because recomputing
# them for every point on the graph for every graph request
# becomes prohitively expensive. Only a few specific rates
# are "recorded" and the ``interval`` parameter must match one
# of those. :(
#
# See prometheus.yaml for the recording rules.
expr="""
100
* cpu:container_usage_seconds:rate{}
/ cores:machine_cpu:total
""".format(interval),
legendFormat="CPU Usage ({} avg)".format(interval),
refId=refId(n),
)
for n, interval in enumerate(intervals),
),
)
def Graph(
id, title, targets, dashLength=None, dashes=False, spaceLength=None,
xAxis=None, yAxes=None, nullPointMode='connected',
):
def merge_target(target):
return {**{
'intervalFactor': 2,
'legendFormat': '',
'refId': 'A',
'step': 600,
}, **target}
targets = [merge_target(t) for t in targets]
assert isinstance(yAxes, YAxes)
return core.Graph(
id=id, title=title, dashLength=dashLength, dashes=dashes,
spaceLength=spaceLength, targets=targets, xAxis=xAxis, yAxes=yAxes,
dataSource='prometheus', nullPointMode=nullPointMode, editable=False,
)
def test_panel_extra_json():
data_source = 'dummy data source'
targets = ['dummy_prom_query']
title = 'dummy title'
extraJson = {
'fillGradient': 6,
'yaxis': {'align': True},
'legend': {'avg': True},
}
graph = G.Graph(data_source, targets, title, extraJson=extraJson)
data = graph.to_json_data()
assert data['targets'] == targets
assert data['datasource'] == data_source
assert data['title'] == title
assert 'alert' not in data
assert data['fillGradient'] == 6
assert data['yaxis']['align'] is True
# Nested non-dict object should also be deep-updated
assert data['legend']['max'] is False
assert data['legend']['avg'] is True
def test_serialization():
"""Serializing a graph doesn't explode."""
graph = G.Graph(
title="CPU Usage by Namespace (rate[5m])",
dataSource="My data source",
targets=[
G.Target(
expr='namespace:container_cpu_usage_seconds_total:sum_rate',
legendFormat='{{namespace}}',
refId='A',
),
],
id=1,
yAxes=[
G.YAxis(format=G.SHORT_FORMAT, label="CPU seconds / second"),
G.YAxis(format=G.SHORT_FORMAT),
],
)
stream = StringIO()
_gen.write_dashboard(graph, stream)
assert stream.getvalue() != ''
def test_alert_thresholds():
some_target_for_alert = G.GraphiteTarget(refId="A", target="foo.bar")
graph = G.Graph(title="Graph with alert",
targets=[some_target_for_alert],
alert=G.Alert(name="alert name",
message="alert message",
alertConditions=[
G.AlertCondition(
some_target_for_alert,
timeRange=G.TimeRange("5m", "now"),
evaluator=G.GreaterThan(10),
reducerType=G.RTYPE_MAX,
operator=G.OP_AND)
]))
stream = StringIO()
_gen.write_dashboard(graph, stream)
graph_json = json.loads(stream.getvalue())
print(graph_json.keys())
#threre is a threshold
assert graph_json['thresholds'][0] != None
def test_auto_refids():
"""
auto_ref_ids() provides refIds for all targets without refIds already
set.
"""
dashboard = G.Dashboard(
title="Test dashboard",
rows=[
G.Row(panels=[
G.Graph(
title="CPU Usage by Namespace (rate[5m])",
targets=[G.Target(expr="metric %d" % i)
for i in range(53)],
).auto_ref_ids()
]),
],
)
assert dashboard.rows[0].panels[0].targets[0].refId == 'A'
assert dashboard.rows[0].panels[0].targets[25].refId == 'Z'
assert dashboard.rows[0].panels[0].targets[26].refId == 'AA'
assert dashboard.rows[0].panels[0].targets[51].refId == 'AZ'
assert dashboard.rows[0].panels[0].targets[52].refId == 'BA'
def PromGraph(title, expressions, **kwargs):
"""Create a graph that renders Prometheus data.
:param title: The title of the graph.
:param expressions: List of tuples of (legend, expr), where 'expr' is a
Prometheus expression.
:param kwargs: Passed on to Graph.
"""
letters = string.ascii_uppercase
expressions = list(expressions)
if len(expressions) > len(letters):
raise ValueError(
'Too many expressions. Can support at most {}, but got {}'.format(
len(letters), len(expressions)))
targets = [
G.Target(expr, legend, refId=refId)
for ((legend, expr), refId) in zip(expressions, letters)
]
return G.Graph(title=title,
dataSource=PROMETHEUS,
targets=targets,
**kwargs)
def test_serialization_zabbix_target():
"""Serializing a graph doesn't explode."""
graph = G.Graph(
title="CPU Usage",
dataSource="Zabbix data source",
targets=[
Z.zabbixMetricTarget(group="Zabbix Group",
host="Zabbix Host",
application="CPU",
item="/CPU (load)/",
functions=[
Z.ZabbixSetAliasFunction("View alias"),
]),
],
id=1,
yAxes=[
G.YAxis(format=G.SHORT_FORMAT, label="CPU seconds / second"),
G.YAxis(format=G.SHORT_FORMAT),
],
)
stream = StringIO()
_gen.write_dashboard(graph, stream)
assert stream.getvalue() != ''
def network_usage(datasource):
return G.Graph(
title="Network Usage",
dataSource=datasource,
xAxis=X_TIME,
yAxes=[
G.YAxis(
# 2^20 bytes / second
format="MBs",
label="Transferred",
),
G.YAxis(
show=False,
),
],
targets=[
G.Target(
# Get the rate of data received on the public interface (eth0)
# for each entire node (id="/") over the last minute.
expr="""
receive:container_network_bytes:rate1m / 2 ^ 20
""",
legendFormat="receive",
refId="A",
),
G.Target(
# And rate of data sent.
expr="""
transmit:container_network_bytes:rate1m / 2 ^ 20
""",
legendFormat="transmit",
refId="B",
),
],
)
def test_serialization_opentsdb_target():
"""Serializing a graph doesn't explode."""
graph = G.Graph(
title="CPU Usage",
dataSource="OpenTSDB data source",
targets=[
O.OpenTSDBTarget(metric='cpu',
alias='$tag_instance',
filters=[
O.OpenTSDBFilter(value='*',
tag='instance',
type='wildcard',
groupBy=True),
]),
],
id=1,
yAxes=[
G.YAxis(format=G.SHORT_FORMAT, label="CPU seconds / second"),
G.YAxis(format=G.SHORT_FORMAT),
],
)
stream = StringIO()
_gen.write_dashboard(graph, stream)
assert stream.getvalue() != ''