class TestMetric(unittest.TestCase):
def setUp(self):
self.dp1 = Metric("/tmp/dp", "test.cpu")
self.dp2 = Metric("/tmp/dp", "test.memory")
def tearDown(self):
self.dp1.purge()
self.dp2.purge()
def test_validate_archives(self):
self.assertTrue(whisper.validateArchiveList(self.dp1.archives))
def test_update(self):
start = int(time.time())
for i in range(start - 1000, start, 1):
self.dp1.update(random.randint(0, 100), i)
data_range, values = self.dp1.fetch(start - 1000, start)
self.assertEqual(1000, len(values))
self.assertEqual(data_range[0], start - 999)
self.assertEqual(data_range[1], start + 1)
def test_fetch(self):
points = [random.randint(0, 100) for i in range(0, 1000)]
start = int(time.time())
index = 0
for i in range(start - 1000, start, 1):
self.dp1.update(points[index], i)
index += 1
data_range, values = self.dp1.fetch(start - 1000, start)
self.assertEqual(points[1:], values[0:-1])
self.assertEqual(sum(points[1:]), sum(values[0:-1]))
class TestMetric(unittest.TestCase):
def setUp(self):
self.dp1 = Metric("/tmp/dp", "test.cpu")
self.dp2 = Metric("/tmp/dp", "test.memory")
def tearDown(self):
self.dp1.purge()
self.dp2.purge()
def test_validate_archives(self):
self.assertTrue(whisper.validateArchiveList(self.dp1.archives))
def test_update(self):
start = int(time.time())
for i in range(start - 1000, start, 1):
self.dp1.update(random.randint(0, 100), i)
data_range, values = self.dp1.fetch(start - 1000, start)
self.assertEqual(1000, len(values))
self.assertEqual(data_range[0], start - 999)
self.assertEqual(data_range[1], start + 1)
def test_fetch(self):
points = [random.randint(0, 100) for i in range(0, 1000)]
start = int(time.time())
index = 0
for i in range(start - 1000, start, 1):
self.dp1.update(points[index], i)
index += 1
data_range, values = self.dp1.fetch(start - 1000, start)
self.assertEqual(points[1:], values[0:-1])
self.assertEqual(sum(points[1:]), sum(values[0:-1]))
def __init__(self, node_id, config):
"""
Creates a new health object for the given node. Databases for health statistics will be created in
the 'data_dir'. The databases use the "whisper" database format from graphite, so they automatically
handle long-term storage with decreasing resolution. Once the databases are created they never grow
or shrink, regardless of the amount of data stored.
:param node_id: The node to create this health object for.
:param data_dir: The data directory to use for the health stats.
"""
self.path = os.path.join(config.get("metrics", "data_dir"), "health", node_id[0], str(node_id[1]))
self.node_id = node_id
self.memory_alert = False
self.swap_alert = False
self.cpu_load = Metric(self.path, "cpu.load")
self.mem = [Metric(self.path, "mem.%s" % item) for item in memory]
self.swp = [Metric(self.path, "swp.%s" % item) for item in swap]
self.cpu = [Metric(self.path, "cpu.%s" % item) for item in cpu]
self.dsk_sp = {}
self.dsk_io = {}
self.net = {}
class Health(object):
def __init__(self, node_id, config):
"""
Creates a new health object for the given node. Databases for health statistics will be created in
the 'data_dir'. The databases use the "whisper" database format from graphite, so they automatically
handle long-term storage with decreasing resolution. Once the databases are created they never grow
or shrink, regardless of the amount of data stored.
:param node_id: The node to create this health object for.
:param data_dir: The data directory to use for the health stats.
"""
self.path = os.path.join(config.get("metrics", "data_dir"), "health", node_id[0], str(node_id[1]))
self.node_id = node_id
self.memory_alert = False
self.swap_alert = False
self.cpu_load = Metric(self.path, "cpu.load")
self.mem = [Metric(self.path, "mem.%s" % item) for item in memory]
self.swp = [Metric(self.path, "swp.%s" % item) for item in swap]
self.cpu = [Metric(self.path, "cpu.%s" % item) for item in cpu]
self.dsk_sp = {}
self.dsk_io = {}
self.net = {}
def get_metric_names(self):
metrics = []
for root, dirnames, filenames in os.walk(self.path):
for filename in filenames:
if filename.endswith(".dp"):
metrics.append(os.path.join(root, filename).replace(self.path, "").replace("/", ".")[1:-3])
return sorted(metrics)
def capture(self):
"""
Captures stats of the local system and writes them into the series database.
:return: None
"""
self.cpu_load.update(psutil.cpu_percent(interval=None))
for i, value in enumerate(psutil.cpu_times()):
self.cpu[i].update(value)
for i,value in enumerate(psutil.virtual_memory()):
self.mem[i].update(value)
for i,value in enumerate(psutil.swap_memory()):
self.swp[i].update(value)
net_io_data = psutil.net_io_counters(pernic=True)
for name in net_io_data:
if name not in self.net:
self.net[name] = [Metric(self.path, "net.io.%s.%s" % (name,item)) for item in net_io]
net = self.net[name]
for i,value in enumerate(net_io_data[name]):
net[i].update(value)
dsk_io_data = psutil.disk_io_counters(perdisk=True)
for name in dsk_io_data:
if name not in self.dsk_io:
self.dsk_io[name] = [Metric(self.path, "dsk.io.%s.%s" % (name,item)) for item in disk_io]
dsk_io = self.dsk_io[name]
for i,value in enumerate(dsk_io_data[name]):
dsk_io[i].update(value)
self.disk_partitions = psutil.disk_partitions()
for disks in self.disk_partitions:
device = disks[0].replace("/dev/", "")
name = "-".join([el for el in device.split("/") if el])
# Create an new set of data points if we find a new disk.
if name not in self.dsk_sp:
self.dsk_sp[name] = [Metric(self.path, "dsk.space.%s.%s" % (name,item)) for item in disk_space]
# Find the disk we are storing data for
dsk = self.dsk_sp[name]
# Update the disk stats
for i, value in enumerate(psutil.disk_usage(disks[1])):
dsk[i].update(value)
def lookup(self, name):
"""
Lookup a metric name and resolve it to a metric database.
:param name: The metric name to resolve.
:return: A data point if it was resolvable, or None
"""
parts = name.split(".")
if parts[0] == "cpu":
if parts[1] == "load":
return self.cpu_load
return self.cpu[cpu.index(parts[1])]
elif parts[0] == "mem":
return self.mem[memory.index(parts[1])]
elif parts[0] == "dsk":
if parts[1] == "space":
return self.dsk_sp[parts[2]][disk_space.index(parts[3])]
elif parts[1] == "io":
return self.dsk_io[parts[2]][disk_io.index(parts[3])]
elif parts[0] == "net":
if parts[1] == "io":
return self.net_io[parts[2]][net_io.index(parts[3])]
return None
def min(self, dp, from_time, until_time=None):
"""
Request the minimum value from the given metric.
:param dp: The metric to check for minimum value.
:param from_time: The earliest time in the series.
:param until_time: The latest time in the series (optional). If omitted this defaults to now.
:return: The minimum value from the series requested.
"""
if type(dp) == type(str()):
dp = self.lookup(dp)
return min([x for x in dp.fetch(from_time, until_time)[1] if x is not None])
def max(self, dp, from_time, until_time=None):
"""
Request the maximum value from the given metric.
:param dp: The metric to check for maximum value.
:param from_time: The earliest time in the series.
:param until_time: The latest time in the series (optional). If omitted this defaults to now.
:return: The maximum value from the series requested.
"""
if type(dp) == type(str()):
dp = self.lookup(dp)
return max([x for x in dp.fetch(from_time, until_time)[1] if x is not None])
def avg(self, dp, from_time, until_time=None):
"""
Request the average value for the given metric.
:param dp: The metric to use to compute the average value.
:param from_time: The earliest time in the series.
:param until_time: The latest time in the series (optional). If omitted this defaults to now.
:return: The average value from the series requested.
"""
if type(dp) == type(str()):
dp = self.lookup(dp)
values = [x for x in dp.fetch(from_time, until_time)[1] if x is not None]
return sum(values) / len(values)
def is_healthy(self, dp, seconds, has_alert, low_water, high_water):
"""
Checks to see if the given metric has been healthy over the last 'seconds' seconds. If 'has_alert' is true then
the metric must be lower than 'low_water', otherwise it must be lower than 'high_water'. Returns True if it's
healthy, false if it's not.
:param dp: The metric to check.
:param seconds: The number of seconds of history to evaluate.
:param has_alert: True if the metric was previously in an unhealthy state.
:param low_water: The low water mark if has_alert is True.
:param high_water: The high water mark.
:return: True if the metric is healthy, False otherwise.
"""
percent_used = self.avg(dp, time.time() - seconds)
if has_alert:
return percent_used < low_water
return percent_used < high_water
def is_memory_healthy(self, seconds, low_water, high_water):
"""
Checks to see if memory is in a healthy state. This is a convenience for is_healthy("mem.percent")
:param seconds: The number of seconds of history to check for health.
:param low_water: The low water level in memory percent used.
:param high_water: The high water level in memory percent used.
:return: True if memory is healthy, False otherwise.
"""
self.memory_alert = not self.is_healthy("mem.percent", seconds, self.memory_alert, low_water, high_water)
return not self.memory_alert
def is_swap_healthy(self, seconds, low_water, high_water):
"""
Checks to see if swap is in a healthy state. This is a convenience for is_healthy("swp.percent")
:param seconds: The number of seconds of history to check for health.
:param low_water: The low water level in swap percent used.
:param high_water: The high water level in swap percent used.
:return: True if swap is healthy, False otherwise.
"""
self.swap_alert = not self.is_healthy("swp.percent", seconds, self.swap_alert, low_water, high_water)
return not self.swap_alert
def setUp(self):
self.dp1 = Metric("/tmp/dp", "test.cpu")
self.dp2 = Metric("/tmp/dp", "test.memory")
def setUp(self):
self.dp1 = Metric("/tmp/dp", "test.cpu")
self.dp2 = Metric("/tmp/dp", "test.memory")
