def deserialize(self, result):
"""
Based on configuration, either stream-deserialize a response in settings.REMOTE_BUFFER_SIZE chunks,
or read the entire payload and use inline deserialization.
:param result: an http response object
:return: deserialized response payload from cluster server
"""
start = time.time()
try:
should_buffer = settings.REMOTE_BUFFER_SIZE > 0
measured_reader = MeasuredReader(BufferedHTTPReader(result, settings.REMOTE_BUFFER_SIZE))
if should_buffer:
log.debug("Using streaming deserializer.")
reader = BufferedHTTPReader(measured_reader, settings.REMOTE_BUFFER_SIZE)
return self._deserialize_stream(reader, result.getheader('content-type'))
log.debug("Using inline deserializer for small payload")
return self._deserialize_buffer(measured_reader.read(), result.getheader('content-type'))
except Exception as err:
self.fail()
log.exception(
"RemoteFinder[%s] Error decoding response from %s: %s" %
(self.host, result.url_full, err))
raise Exception("Error decoding response from %s: %s" % (result.url_full, err))
finally:
log.debug("Processed %d bytes in %f seconds." % (measured_reader.bytes_read, time.time() - start))
result.release_conn()
def set_metadata_view(request):
results = {}
if request.method == 'GET':
metric = request.GET['metric']
key = request.GET['key']
value = request.GET['value']
try:
results[metric] = CarbonLink.set_metadata(metric, key, value)
except:
log.exception()
results[metric] = dict(error="Unexpected error occurred in CarbonLink.set_metadata(%s, %s)" % (metric, key))
elif request.method == 'POST':
if request.META.get('CONTENT_TYPE') == 'application/json':
operations = json.loads( request.body )
else:
operations = json.loads( request.POST['operations'] )
for op in operations:
metric = None
try:
metric, key, value = op['metric'], op['key'], op['value']
results[metric] = CarbonLink.set_metadata(metric, key, value)
except:
log.exception()
if metric:
results[metric] = dict(error="Unexpected error occurred in bulk CarbonLink.set_metadata(%s)" % metric)
else:
results = dict(error="Invalid request method")
return json_response_for(request, results)
def get_results(self):
if self.failed:
return
if self.cachedResult is not None:
results = self.cachedResult
else:
if self.connection is None:
self.send()
try:
response = self.connection.getresponse()
assert response.status == 200, "received error response %s - %s" % (response.status, response.reason)
result_data = response.read()
results = unpickle.loads(result_data)
except:
log.exception("FindRequest.get_results(host=%s, query=%s) exception processing response" % (self.store.host, self.query))
self.store.fail()
return
cache.set(self.cacheKey, results, settings.FIND_CACHE_DURATION)
for node_info in results:
if node_info.get('is_leaf'):
reader = RemoteReader(self.store, node_info, bulk_query=self.query.pattern)
node = LeafNode(node_info['path'], reader)
else:
node = BranchNode(node_info['path'])
node.local = False
yield node
def _dosave(request,viewName):
profile = getProfile(request)
#First find our View
log.info("Saving view '%s' under profile '%s'" % (viewName,profile.user.username))
try:
view = profile.view_set.get(name=viewName)
except ObjectDoesNotExist:
view = View(profile=profile,name=viewName)
view.save()
#Now re-associate the view with the correct Windows
view.window_set.all().delete()
for windowName,encodedString in request.GET.items():
try:
if windowName in ('_','commandInput'): continue
paramString = urllib.unquote_plus(encodedString)
queryParams = cgi.parse_qs(paramString)
modelParams = {}
for key,value in queryParams.items(): #Clean up the window params
key = str(key)
value = str(value[0])
if key in ('top','left'):
value = int(float( value.replace('px','') ))
if key in ('width','height','interval'):
value = int(float(value))
modelParams[key] = value
if 'interval' not in modelParams:
modelParams['interval'] = None
win = Window(view=view,name=windowName,**modelParams)
win.save()
except:
log.exception("Failed to process parameters for window '%s'" % windowName)
return stdout('Saved view %s' % viewName)
def fetch(self, startTime, endTime):
data = whisper.fetch(self.fs_path, startTime, endTime)
if not data:
return None
time_info, values = data
(start,end,step) = time_info
# Merge in data from carbon's cache
try:
cached_datapoints = CarbonLink.query(self.real_metric_path)
except:
log.exception("Failed CarbonLink query '%s'" % self.real_metric_path)
cached_datapoints = []
for (timestamp, value) in cached_datapoints:
interval = timestamp - (timestamp % step)
try:
i = int(interval - start) / step
values[i] = value
except:
pass
return (time_info, values)
def evaluateTokens(requestContext, tokens):
if tokens.expression:
return evaluateTokens(requestContext, tokens.expression)
elif tokens.pathExpression:
return fetchData(requestContext, tokens.pathExpression)
elif tokens.call:
try:
func = SeriesFunctions[tokens.call.func]
args = [evaluateTokens(requestContext, arg) for arg in tokens.call.args]
return func(requestContext, *args)
except ValueError:
log.exception('value error when render')
return []
elif tokens.number:
if tokens.number.integer:
return int(tokens.number.integer)
elif tokens.number.float:
return float(tokens.number.float)
elif tokens.number.scientific:
return float(tokens.number.scientific[0])
elif tokens.string:
return tokens.string[1:-1]
elif tokens.boolean:
return tokens.boolean[0] == 'true'
def fetch(self, startTime, endTime, now=None, requestContext=None):
# Start the fetch on each node
fetches = []
for n in self.nodes:
try:
fetches.append(n.fetch(startTime, endTime, now, requestContext))
except:
log.exception("Failed to initiate subfetch for %s" % str(n))
def merge_results():
results = {}
# Wait for any asynchronous operations to complete
for i, result in enumerate(fetches):
if isinstance(result, FetchInProgress):
try:
results[i] = result.waitForResults()
except:
log.exception("Failed to complete subfetch")
results[i] = None
else:
results[i] = result
results = [r for r in results.values() if r is not None]
if not results:
raise Exception("All sub-fetches failed")
return reduce(self.merge, results)
return FetchInProgress(merge_results)
def request_series():
if request_lock.acquire(False): # the FetchInProgress that gets waited on waits for the actual completion
try:
log.info("RemoteReader.request_data :: requesting %s" % url)
connection = HTTPConnectionWithTimeout(self.store.host)
connection.timeout = settings.REMOTE_FETCH_TIMEOUT
connection.request('GET', urlpath)
response = connection.getresponse()
if response.status != 200:
raise Exception("Error response %d %s from %s" % (response.status, response.reason, url))
pickled_response = response.read()
results = unpickle.loads(pickled_response)
self.cache_lock.acquire()
self.request_cache[url] = results
self.cache_lock.release()
completion_event.set()
return results
except:
completion_event.set()
self.store.fail()
log.exception("Error requesting %s" % url)
raise
else: # otherwise we just wait on the completion_event
completion_event.wait(settings.REMOTE_FETCH_TIMEOUT)
cached_results = self.request_cache.get(url)
if cached_results is None:
raise Exception("Passive remote fetch failed to find cached results")
else:
return cached_results
def fetch(self, startTime, endTime):
data = whisper.fetch(self.fs_path, startTime, endTime)
if not data:
return None
time_info, values = data
(start, end, step) = time_info
meta_info = whisper.info(self.fs_path)
lowest_step = min([i['secondsPerPoint'] for i in meta_info['archives']])
# Merge in data from carbon's cache
cached_datapoints = []
try:
if step == lowest_step:
cached_datapoints = CarbonLink.query(self.real_metric_path)
except:
log.exception("Failed CarbonLink query '%s'" % self.real_metric_path)
cached_datapoints = []
if isinstance(cached_datapoints, dict):
cached_datapoints = cached_datapoints.items()
for (timestamp, value) in cached_datapoints:
interval = timestamp - (timestamp % step)
try:
i = int(interval - start) / step
values[i] = value
except:
pass
return (time_info, values)
def auto_complete_values(self, exprs, tag, valuePrefix=None, limit=None, requestContext=None):
"""
Return auto-complete suggestions for tags and values based on the matches for the specified expressions, optionally filtered by tag and/or value prefix
"""
if limit is None:
limit = settings.TAGDB_AUTOCOMPLETE_LIMIT
fields = [
('tag', tag or ''),
('valuePrefix', valuePrefix or ''),
('limit', str(limit)),
('local', self.params.get('local', '1')),
]
for expr in exprs:
fields.append(('expr', expr))
result = self.request(
'/tags/autoComplete/values',
fields,
headers=requestContext.get('forwardHeaders') if requestContext else None,
timeout=settings.FIND_TIMEOUT)
try:
reader = codecs.getreader('utf-8')
results = json.load(reader(result))
except Exception as err:
self.fail()
log.exception(
"RemoteFinder[%s] Error decoding autocomplete values response from %s: %s" %
(self.host, result.url_full, err))
raise Exception("Error decoding autocomplete values response from %s: %s" % (result.url_full, err))
finally:
result.release_conn()
return results
def wait_for_results():
connection_event.wait(1)
connection = self.request_connections.get(url)
if not connection:
log.exception("RemoteReader.wait_for_results :: no connection found")
if wait_lock.acquire(False): # the FetchInProgress that gets waited on waits for the actual completion
try:
response = connection.getresponse()
if response.status != 200:
raise Exception("Error response %d %s from %s" % (response.status, response.reason, url))
pickled_response = response.read()
results = unpickle.loads(pickled_response)
self.cache_lock.acquire()
self.request_cache[url] = results
self.cache_lock.release()
completion_event.set()
return results
except:
completion_event.set()
self.store.fail()
log.exception("Error requesting %s" % url)
raise
else: # otherwise we just wait on the completion_event
completion_event.wait(settings.REMOTE_FETCH_TIMEOUT)
cached_results = self.request_cache.get(url)
if cached_results is None:
raise Exception("Passive remote fetch failed to find cached results")
else:
return cached_results
def delegateRenderIOW(graphOptions,graphType,tenant):
start = time()
log.rendering(graphOptions['data'])
post_data = {'target': [], 'from': '-24hours'}
if 'data' in graphOptions:
for series in graphOptions['data']:
log.rendering(series.name)
post_data['target'].append(series.name)
post_data['from'] = series.start
post_data['until'] = series.end
post_data['tenant']=tenant
post_data['graphType']=graphType
servers = settings.RENDERING_HOSTS[:] #make a copy so we can shuffle it safely
shuffle(servers)
for server in servers:
start2 = time()
try:
response = requests.post("%s/render/" % server, data=post_data)
assert response.status_code == 200, "Bad response code %d from %s" % (response.status_code,server)
contentType = response.headers['Content-Type']
imageData = response.content
assert contentType == 'image/png', "Bad content type: \"%s\" from %s" % (contentType,server)
assert imageData, "Received empty response from %s" % server
# Wrap things up
log.rendering('Remotely rendered image on %s in %.6f seconds' % (server,time() - start2))
log.rendering('Spent a total of %.6f seconds doing remote rendering work' % (time() - start))
return imageData
except:
log.exception("Exception while attempting remote rendering request on %s" % server)
log.rendering('Exception while remotely rendering on %s wasted %.6f' % (server,time() - start2))
continue
def set_metadata_view(request):
results = {}
if request.method == "GET":
metric = request.GET["metric"]
key = request.GET["key"]
value = request.GET["value"]
try:
results[metric] = CarbonLink.set_metadata(metric, key, value)
except:
log.exception()
results[metric] = dict(error="Unexpected error occurred in CarbonLink.set_metadata(%s, %s)" % (metric, key))
elif request.method == "POST":
if request.META.get("CONTENT_TYPE") == "application/json":
operations = json.loads(request.body)
else:
operations = json.loads(request.POST["operations"])
for op in operations:
metric = None
try:
metric, key, value = op["metric"], op["key"], op["value"]
results[metric] = CarbonLink.set_metadata(metric, key, value)
except:
log.exception()
if metric:
results[metric] = dict(
error="Unexpected error occurred in bulk CarbonLink.set_metadata(%s)" % metric
)
else:
results = dict(error="Invalid request method")
return json_response_for(request, results)
def send(self):
log.info("FindRequest.send(host=%s, query=%s) called" % (self.store.host, self.query))
self.cachedResult = cache.get(self.cacheKey)
if self.cachedResult is not None:
log.info("FindRequest(host=%s, query=%s) using cached result" % (self.store.host, self.query))
return
self.connection = HTTPConnectionWithTimeout(self.store.host)
self.connection.timeout = settings.REMOTE_FIND_TIMEOUT
query_params = [("local", "1"), ("format", "pickle"), ("query", self.query.pattern)]
if self.query.startTime:
query_params.append(("from", self.query.startTime))
if self.query.endTime:
query_params.append(("until", self.query.endTime))
query_string = urlencode(query_params)
try:
self.connection.request("GET", "/metrics/find/?" + query_string)
except:
log.exception(
"FindRequest.send(host=%s, query=%s) exception during request" % (self.store.host, self.query)
)
self.store.fail()
self.failed = True
def is_local_interface(host):
is_ipv6 = False
if ':' in host:
try:
if host.find('[', 0, 2) != -1:
last_bracket_position = host.rfind(']')
last_colon_position = host.rfind(':')
if last_colon_position > last_bracket_position:
host = host.rsplit(':', 1)[0]
host = host.strip('[]')
socket.inet_pton(socket.AF_INET6, host)
is_ipv6 = True
except socket.error:
host = host.split(':',1)[0]
try:
if is_ipv6:
sock = socket.socket(socket.AF_INET6, socket.SOCK_DGRAM)
else:
sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
sock.connect( (host, 4242) )
local_ip = sock.getsockname()[0]
sock.close()
except:
log.exception("Failed to open socket with %s" % host)
raise
if local_ip == host:
return True
return False
def request(self, path, fields=None, headers=None, timeout=None):
url = "%s%s" % (self.url, path)
url_full = "%s?%s" % (url, urlencode(fields))
try:
result = http.request(
'POST' if settings.REMOTE_STORE_USE_POST else 'GET',
url,
fields=fields,
headers=headers,
timeout=timeout,
preload_content=False)
except BaseException as err:
self.fail()
log.exception("RemoteFinder[%s] Error requesting %s: %s" % (self.host, url_full, err))
raise Exception("Error requesting %s: %s" % (url_full, err))
if result.status != 200:
result.release_conn()
self.fail()
log.exception(
"RemoteFinder[%s] Error response %d from %s" % (self.host, result.status, url_full))
raise Exception("Error response %d from %s" % (result.status, url_full))
result.url_full = url_full
# reset last failure time so that retried fetches can re-enable a remote
self.last_failure = 0
log.debug("RemoteFinder[%s] Fetched %s" % (self.host, url_full))
return result
def fetch(self, startTime, endTime):
data = whisper.fetch(self.fs_path, startTime, endTime)
if not data:
return None
consolidationFunc = ""
whisper_info = whisper.info(self.fs_path)
if "aggregationMethod" in whisper_info:
aggregationMethod = whisper_info["aggregationMethod"]
if aggregationMethod == 'min' or aggregationMethod == 'max':
consolidationFunc = aggregationMethod
time_info, values = data
(start,end,step) = time_info
# Merge in data from carbon's cache
try:
cached_datapoints = CarbonLink.query(self.real_metric_path)
except:
log.exception("Failed CarbonLink query '%s'" % self.real_metric_path)
cached_datapoints = []
for (timestamp, value) in cached_datapoints:
interval = timestamp - (timestamp % step)
try:
i = int(interval - start) / step
values[i] = value
except:
pass
return (time_info, values, consolidationFunc)
def receiveResponse():
try:
buf = ''
remaining = 4
message_size = None
while remaining:
packet = connection.recv(remaining)
assert packet, "CarbonLink lost connection to %s:%d" % host
buf += packet
if message_size is None:
if len(buf) == 4:
remaining = message_size = struct.unpack("!L", buf)[0]
buf = ''
continue
remaining -= len(packet)
# We're done with the connection for this request, put it in the pool
self.putConnectionInPool(host, connection)
# Now parse the response
points = pickle.loads(buf)
log.cache("CarbonLink to %s, retrieved %d points for %s" % (host,len(points),metric))
for point in points:
yield point
except:
log.exception("CarbonLink to %s, exception while getting response" % str(host))
self.removeConnectionFromPool(host, connection)
def fetchDataLocal(requestContext, pathExpr):
seriesList = []
startTime = requestContext['startTime']
endTime = requestContext['endTime']
if requestContext['localOnly']:
store = LOCAL_STORE
else:
store = STORE
for dbFile in store.find(pathExpr):
log.metric_access(dbFile.metric_path)
dbResults = dbFile.fetch( timestamp(startTime), timestamp(endTime) )
try:
cachedResults = CarbonLink.query(dbFile.real_metric)
results = mergeResults(dbResults, cachedResults)
except:
log.exception()
results = dbResults
if not results:
continue
(timeInfo,values) = results
(start,end,step) = timeInfo
series = TimeSeries(dbFile.metric_path, start, end, step, values)
series.pathExpression = pathExpr #hack to pass expressions through to render functions
seriesList.append(series)
return seriesList
def fetch(self, startTime, endTime):
try:
data = whisper.fetch(self.fs_path, startTime, endTime)
except IOError:
log.exception("Failed fetch of whisper file '%s'" % self.fs_path)
return None
if not data:
return None
time_info, values = data
(start,end,step) = time_info
meta_info = whisper.info(self.fs_path)
aggregation_method = meta_info['aggregationMethod']
lowest_step = min([i['secondsPerPoint'] for i in meta_info['archives']])
# Merge in data from carbon's cache
cached_datapoints = []
try:
cached_datapoints = CarbonLink().query(self.real_metric_path)
except:
log.exception("Failed CarbonLink query '%s'" % self.real_metric_path)
cached_datapoints = []
if isinstance(cached_datapoints, dict):
cached_datapoints = cached_datapoints.items()
values = merge_with_cache(cached_datapoints,
start,
step,
values,
aggregation_method)
return time_info, values
def mygraph(request):
profile = getProfile(request,allowDefault=False)
if not profile: return HttpResponse( "You are not logged in!" )
graphName = request.GET['graphName']
if not graphName:
return HttpResponse("You must type in a graph name.")
action = request.GET['action']
url = request.GET['url']
if action == 'save':
try:
existingGraph = profile.mygraph_set.get(name=graphName)
existingGraph.url = url
existingGraph.save()
except ObjectDoesNotExist:
try:
newGraph = MyGraph(profile=profile,name=graphName,url=url)
newGraph.save()
except:
log.exception("Failed to create new MyGraph in /composer/mygraph/, graphName=%s" % graphName)
return HttpResponse("Failed to save graph %s" % graphName)
return HttpResponse("SAVED")
elif action == 'delete':
try:
existingGraph = profile.mygraph_set.get(name=graphName)
existingGraph.delete()
except ObjectDoesNotExist:
return HttpResponse("No such graph '%s'" % graphName)
return HttpResponse("DELETED")
else:
return HttpResponse("Invalid operation '%s'" % action)
def get_index(self, requestContext):
url = '/metrics/index.json'
headers = requestContext.get('forwardHeaders')
result = self.request(
url,
fields=[
('local', self.params.get('local', '1')),
],
headers=headers,
timeout=settings.FIND_TIMEOUT)
try:
reader = codecs.getreader('utf-8')
results = json.load(reader(result))
except Exception as err:
self.fail()
log.exception(
"RemoteFinder[%s] Error decoding index response from %s: %s" %
(self.host, result.url_full, err))
raise Exception("Error decoding index response from %s: %s" % (result.url_full, err))
finally:
result.release_conn()
return results
def renderView(request):
start = time()
(graphOptions, requestOptions) = parseOptions(request)
post_data = {
'from' : int((requestOptions['startTime']-datetime(1970,1,1,tzinfo=requestOptions['tzinfo'])).total_seconds()),
'until' : int((requestOptions['endTime']-datetime(1970,1,1,tzinfo=requestOptions['tzinfo'])).total_seconds()),
'tenant' : requestOptions['tenant'],
'format' : requestOptions.get('format'),
'target': [t for t in requestOptions['targets'] if t.strip()],
}
post_data.update(graphOptions)
log.rendering(post_data)
servers = settings.RENDERING_HOSTS[:] #make a copy so we can shuffle it safely
shuffle(servers)
for server in servers:
start2 = time()
try:
response = requests.post("%s/render/" % server, data=post_data)
assert response.status_code == 200, "Bad response code %d from %s" % (response.status_code,server)
contentType = response.headers['Content-Type']
imageData = response.content
assert contentType == 'image/png', "Bad content type: \"%s\" from %s" % (contentType,server)
assert imageData, "Received empty response from %s" % server
# Wrap things up
log.rendering('Remotely rendered image on %s in %.6f seconds' % (server,time() - start2))
log.rendering('Spent a total of %.6f seconds doing remote rendering work' % (time() - start))
except:
log.exception("Exception while attempting remote rendering request on %s" % server)
log.rendering('Exception while remotely rendering on %s wasted %.6f' % (server,time() - start2))
continue
response = buildResponse(imageData, 'image/png')
log.rendering('Total rendering time %.6f seconds' % (time() - start))
return response
def send(self):
log.info("FindRequest.send(host=%s, query=%s) called" % (self.store.host, self.query))
self.cachedResult = cache.get(self.cacheKey)
if self.cachedResult is not None:
log.info("FindRequest(host=%s, query=%s) using cached result" % (self.store.host, self.query))
return
self.connection = HTTPConnectionWithTimeout(self.store.host)
self.connection.timeout = settings.REMOTE_FIND_TIMEOUT
query_params = [
('local', '1'),
('format', 'pickle'),
('query', self.query.pattern),
]
if self.query.startTime:
query_params.append( ('from', self.query.startTime) )
if self.query.endTime:
query_params.append( ('until', self.query.endTime) )
query_string = urlencode(query_params)
try:
self.connection.request('GET', '/metrics/find/?' + query_string)
except:
log.exception("FindRequest.send(host=%s, query=%s) exception during request" % (self.store.host, self.query))
self.store.fail()
self.failed = True
def find_nodes(self, query):
log.info("q:" + repr(query))
try:
for node in self._find_nodes_from_pattern(self.kudu_table, query.pattern):
yield node
except Exception, e:
log.exception(e)
raise
def index_json(request):
queryParams = request.GET.copy()
queryParams.update(request.POST)
jsonp = queryParams.get('jsonp', False)
cluster = queryParams.get('cluster', False)
def find_matches():
matches = []
for root, dirs, files in os.walk(settings.WHISPER_DIR):
root = root.replace(settings.WHISPER_DIR, '')
for basename in files:
if fnmatch.fnmatch(basename, '*.wsp'):
matches.append(os.path.join(root, basename))
for root, dirs, files in os.walk(settings.CERES_DIR):
root = root.replace(settings.CERES_DIR, '')
for filename in files:
if filename == '.ceres-node':
matches.append(root)
# unlike 0.9.x, we're going to use os.walk with followlinks
# since we require Python 2.7 and newer that supports it
if RRDReader.supported:
for root, dirs, files in os.walk(settings.RRD_DIR, followlinks=True):
root = root.replace(settings.RRD_DIR, '')
for basename in files:
if fnmatch.fnmatch(basename, '*.rrd'):
absolute_path = os.path.join(settings.RRD_DIR, root, basename)
(basename,extension) = os.path.splitext(basename)
metric_path = os.path.join(root, basename)
rrd = RRDReader(absolute_path, metric_path)
for datasource_name in rrd.get_datasources(absolute_path):
matches.append(os.path.join(metric_path, datasource_name))
matches = [
m
.replace('.wsp', '')
.replace('.rrd', '')
.replace('/', '.')
.lstrip('.')
for m in sorted(matches)
]
return matches
matches = []
if cluster and len(settings.CLUSTER_SERVERS) >= 1:
try:
matches = reduce( lambda x, y: list(set(x + y)), \
[json.loads(urllib.urlopen('http://' + cluster_server + '/metrics/index.json').read()) \
for cluster_server in settings.CLUSTER_SERVERS])
except urllib.URLError:
log.exception()
return json_response_for(request, matches, jsonp=jsonp, status=500)
else:
matches = find_matches()
return json_response_for(request, matches, jsonp=jsonp)
def fetchData(requestContext, pathExpr):
seriesList = []
startTime = int( time.mktime( requestContext['startTime'].timetuple() ) )
endTime = int( time.mktime( requestContext['endTime'].timetuple() ) )
def _fetchData(pathExpr,startTime, endTime, requestContext, seriesList):
matching_nodes = STORE.find(pathExpr, startTime, endTime, local=requestContext['localOnly'])
fetches = [(node, node.fetch(startTime, endTime)) for node in matching_nodes if node.is_leaf]
for node, results in fetches:
if isinstance(results, FetchInProgress):
results = results.waitForResults()
if not results:
log.info("render.datalib.fetchData :: no results for %s.fetch(%s, %s)" % (node, startTime, endTime))
continue
try:
(timeInfo, values, conso) = results
except ValueError as e:
raise Exception("could not parse timeInfo/values from metric '%s': %s" % (node.path, e))
(start, end, step) = timeInfo
if conso != "":
series = TimeSeries(node.path, start, end, step, values, conso)
else:
series = TimeSeries(node.path, start, end, step, values)
series.pathExpression = pathExpr #hack to pass expressions through to render functions
seriesList.append(series)
# Prune empty series with duplicate metric paths to avoid showing empty graph elements for old whisper data
names = set([ s.name for s in seriesList ])
for name in names:
series_with_duplicate_names = [ s for s in seriesList if s.name == name ]
empty_duplicates = [ s for s in series_with_duplicate_names if not nonempty(s) ]
if series_with_duplicate_names == empty_duplicates and len(empty_duplicates) > 0: # if they're all empty
empty_duplicates.pop() # make sure we leave one in seriesList
for series in empty_duplicates:
seriesList.remove(series)
return seriesList
retries = 1 # start counting at one to make log output and settings more readable
while True:
try:
seriesList = _fetchData(pathExpr,startTime, endTime, requestContext, seriesList)
return seriesList
except Exception, e:
if retries >= settings.MAX_FETCH_RETRIES:
log.exception("Failed after %i retry! See: %s" % (settings.MAX_FETCH_RETRIES, e))
raise Exception("Failed after %i retry! See: %s" % (settings.MAX_FETCH_RETRIES, e))
else:
log.exception("Got an exception when fetching data! See: %s Will do it again! Run: %i of %i" %
(e, retries, settings.MAX_FETCH_RETRIES))
retries += 1
def _find_paths(self, current_dir, patterns):
"""Recursively generates absolute paths whose components underneath current_dir
match the corresponding pattern in patterns"""
pattern = patterns[0]
patterns = patterns[1:]
has_wildcard = (
pattern.find("{") > -1 or pattern.find("[") > -1 or pattern.find("*") > -1 or pattern.find("?") > -1
)
using_globstar = pattern == "**"
if has_wildcard: # this avoids os.listdir() for performance
try:
entries = os.listdir(current_dir)
except OSError as e:
log.exception(e)
entries = []
else:
entries = [pattern]
if using_globstar:
matching_subdirs = map(operator.itemgetter(0), os.walk(current_dir))
else:
subdirs = [entry for entry in entries if isdir(join(current_dir, entry))]
matching_subdirs = match_entries(subdirs, pattern)
# if this is a terminal globstar, add a pattern for all files in subdirs
if using_globstar and not patterns:
patterns = ["*"]
if len(patterns) == 1 and RRDReader.supported: # the last pattern may apply to RRD data sources
if not has_wildcard:
entries = [pattern + ".rrd"]
files = [entry for entry in entries if isfile(join(current_dir, entry))]
rrd_files = match_entries(files, pattern + ".rrd")
if rrd_files: # let's assume it does
datasource_pattern = patterns[0]
for rrd_file in rrd_files:
absolute_path = join(current_dir, rrd_file)
yield absolute_path + self.DATASOURCE_DELIMITER + datasource_pattern
if patterns: # we've still got more directories to traverse
for subdir in matching_subdirs:
absolute_path = join(current_dir, subdir)
for match in self._find_paths(absolute_path, patterns):
yield match
else: # we've got the last pattern
if not has_wildcard:
entries = [pattern + ".wsp", pattern + ".wsp.gz", pattern + ".rrd"]
files = [entry for entry in entries if isfile(join(current_dir, entry))]
matching_files = match_entries(files, pattern + ".*")
for base_name in matching_files + matching_subdirs:
yield join(current_dir, base_name)
def __get_cached_datapoints(self, stage):
cached_datapoints = []
try:
# TODO: maybe this works with non stage0 now, need to check.
if stage.stage0 and self._carbonlink:
cached_datapoints = self._carbonlink.query(self._metric_name)
except Exception:
log.exception("Failed CarbonLink query '%s'" % self._metric_name)
return cached_datapoints
def wait_for_result(result):
"""Helper to read the various result types."""
if isinstance(result, FetchInProgress):
try:
return result.waitForResults()
except BaseException:
log.exception("Failed to complete subfetch")
return None
else:
return result
def fetch(self, startTime, endTime):
data = self.ceres_node.read(startTime, endTime)
time_info = (data.startTime, data.endTime, data.timeStep)
values = list(data.values)
# Merge in data from carbon's cache
try:
cached_datapoints = CarbonLink.query(self.real_metric_path)
except:
log.exception("Failed CarbonLink query '%s'" % self.real_metric_path)
cached_datapoints = []
values = merge_with_cache(cached_datapoints,
data.startTime,
data.timeStep,
values)
return time_info, values
def renderLocalView(request):
try:
start = time()
reqParams = StringIO(request.body)
graphType = reqParams.readline().strip()
optionsPickle = reqParams.read()
reqParams.close()
graphClass = GraphTypes[graphType]
options = unpickle.loads(optionsPickle)
image = doImageRender(graphClass, options)
log.rendering("Delegated rendering request took %.6f seconds" %
(time() - start))
response = buildResponse(image)
add_never_cache_headers(response)
return response
except:
log.exception("Exception in graphite.render.views.rawrender")
return HttpResponseServerError()
def fetch(self, startTime, endTime):
# Start the fetch on each node
results = [n.fetch(startTime, endTime) for n in self.nodes]
# Wait for any asynchronous operations to complete
for i, result in enumerate(results):
if isinstance(result, FetchInProgress):
try:
results[i] = result.waitForResults()
except:
log.exception("Failed to complete subfetch")
results[i] = None
results = [r for r in results if r is not None]
if not results:
raise Exception("All sub-fetches failed")
return reduce(self.merge, results)
def index_json(request):
queryParams = request.GET.copy()
queryParams.update(request.POST)
try:
jsonp = queryParams.get('jsonp', False)
requestContext = {
'localOnly': int( queryParams.get('local', 0) ),
'forwardHeaders': extractForwardHeaders(request),
}
matches = STORE.get_index(requestContext)
except Exception:
log.exception()
return json_response_for(request, [], jsonp=jsonp, status=500)
return json_response_for(request, matches, jsonp=jsonp)
def mygraph(request):
profile = getProfile(request, allowDefault=False)
if not profile:
return HttpResponse("You are not logged in!")
action = request.GET['action']
graphName = request.GET['graphName']
if not graphName:
return HttpResponse("You must type in a graph name.")
if action == 'save':
url = request.GET['url']
try:
existingGraph = profile.mygraph_set.get(name=graphName)
existingGraph.url = url
existingGraph.save()
except ObjectDoesNotExist:
try:
newGraph = MyGraph(profile=profile, name=graphName, url=url)
newGraph.save()
except Exception:
log.exception(
"Failed to create new MyGraph in /composer/mygraph/, graphName=%s"
% graphName)
return HttpResponse("Failed to save graph %s" % graphName)
return HttpResponse("SAVED")
elif action == 'delete':
try:
existingGraph = profile.mygraph_set.get(name=graphName)
existingGraph.delete()
except ObjectDoesNotExist:
return HttpResponse("No such graph '%s'" % graphName)
return HttpResponse("DELETED")
else:
return HttpResponse("Invalid operation '%s'" % action)
def merge_results():
results = {}
# Wait for any asynchronous operations to complete
for i, result in enumerate(fetches):
if isinstance(result, FetchInProgress):
try:
results[i] = result.waitForResults()
except:
log.exception("Failed to complete subfetch")
results[i] = None
else:
results[i] = result
results = [r for r in results.values() if r is not None]
if not results:
raise Exception("All sub-fetches failed")
return reduce(self.merge, results)
def get_metadata_view(request):
queryParams = request.GET.copy()
queryParams.update(request.POST)
key = queryParams.get('key')
metrics = queryParams.getlist('metric')
jsonp = queryParams.get('jsonp', False)
results = {}
for metric in metrics:
try:
results[metric] = CarbonLink.get_metadata(metric, key)
except:
log.exception()
results[metric] = dict(
error=
"Unexpected error occurred in CarbonLink.get_metadata(%s, %s)"
% (metric, key))
return json_response_for(request, results, jsonp=jsonp)
def read_response(self): # called under self.lock
try:
self.has_done_response_read = True
response = self.connection.getresponse(
) # safe if self.connection.timeout works as advertised
if response.status != 200:
raise Exception("Error response %d %s from http://%s%s" %
(response.status, response.reason,
self.store.host, self.urlpath))
pickled_response = response.read()
self.result = unpickle.loads(pickled_response)
return self.result
except:
self.store.fail()
log.exception("Error requesting http://%s%s" %
(self.store.host, self.urlpath))
raise
finally:
self.done_cb()
def _dogsave(request, graphName):
profile = getProfile(request, allowDefault=False)
if not profile: return stderr("You must be logged in to save graphs")
url = request.GET.get('url')
if not url: return stderr("No url specified!")
try:
existingGraph = profile.mygraph_set.get(name=graphName)
existingGraph.url = url
existingGraph.save()
except ObjectDoesNotExist:
try:
newGraph = MyGraph(profile=profile, name=graphName, url=url)
newGraph.save()
except:
log.exception(
"Failed to create new MyGraph in _dogsave(), graphName=%s" %
graphName)
return stderr("Failed to save graph %s" % graphName)
return stdout("Saved graph %s" % graphName)
def delegateRendering(graphType, graphOptions):
start = time()
postData = graphType + '\n' + pickle.dumps(graphOptions)
servers = settings.RENDERING_HOSTS[:] #make a copy so we can shuffle it safely
shuffle(servers)
connector_class = connector_class_selector(settings.INTRACLUSTER_HTTPS)
for server in servers:
start2 = time()
try:
# Get a connection
try:
pool = connectionPools[server]
except KeyError: #happens the first time
pool = connectionPools[server] = set()
try:
connection = pool.pop()
except KeyError: #No available connections, have to make a new one
connection = connector_class(server)
connection.timeout = settings.REMOTE_RENDER_CONNECT_TIMEOUT
# Send the request
try:
connection.request('POST','/render/local/', postData)
except CannotSendRequest:
connection = connector_class(server) #retry once
connection.timeout = settings.REMOTE_RENDER_CONNECT_TIMEOUT
connection.request('POST', '/render/local/', postData)
# Read the response
response = connection.getresponse()
assert response.status == 200, "Bad response code %d from %s" % (response.status,server)
contentType = response.getheader('Content-Type')
imageData = response.read()
assert contentType == 'image/png', "Bad content type: \"%s\" from %s" % (contentType,server)
assert imageData, "Received empty response from %s" % server
# Wrap things up
log.rendering('Remotely rendered image on %s in %.6f seconds' % (server,time() - start2))
log.rendering('Spent a total of %.6f seconds doing remote rendering work' % (time() - start))
pool.add(connection)
return imageData
except:
log.exception("Exception while attempting remote rendering request on %s" % server)
log.rendering('Exception while remotely rendering on %s wasted %.6f' % (server,time() - start2))
continue
def fetchRemoteData(requestContext, pathExpr, usePrefetchCache=settings.REMOTE_PREFETCH_DATA):
(startTime, endTime, now) = _timebounds(requestContext)
remote_nodes = [ RemoteNode(store, pathExpr, True) for store in STORE.remote_stores ]
# Go through all of the remote_nodes, and launch a remote_fetch for each one.
# Each fetch will take place in its own thread, since it's naturally parallel work.
# Notable: return the 'seriesList' result from each node.fetch into result_queue
# instead of directly from the method. Queue.Queue() is threadsafe.
remote_fetches = []
result_queue = Queue.Queue()
for node in remote_nodes:
need_fetch = True
if usePrefetchCache:
series = prefetchLookup(requestContext, node)
# Will be either:
# []: prefetch done, returned no data. Do not fetch
# seriesList: prefetch done, returned data, do not fetch
# None: prefetch not done, FETCH
if series is not None:
result_queue.put( (node, series) )
need_fetch = False
if need_fetch:
fetch_thread = threading.Thread(target=node.fetch, name=node.store.host,
args=(startTime, endTime, now, result_queue))
fetch_thread.start()
remote_fetches.append(fetch_thread)
# Once the remote_fetches have started, wait for them all to finish. Assuming an
# upper bound of REMOTE_STORE_FETCH_TIMEOUT per thread, this should take about that
# amount of time (6s by default) at the longest. If every thread blocks permanently,
# then this could take a horrible REMOTE_STORE_FETCH_TIMEOUT * num(remote_fetches),
# but then that would imply that remote_storage's HTTPConnectionWithTimeout class isn't
# working correctly :-)
for fetch_thread in remote_fetches:
try:
fetch_thread.join(settings.REMOTE_STORE_FETCH_TIMEOUT)
if fetch_thread.is_alive():
log.exception("Failed to join remote_fetch thread %s within %ss" % (fetch_thread.name, settings.REMOTE_STORE_FETCH_TIMEOUT))
except:
log.exception("Exception during remote_fetch thread %s" % (fetch_thread.name))
return result_queue
def fetchData(requestContext, pathExpr):
seriesList = {}
(startTime, endTime, now) = timebounds(requestContext)
retries = 1 # start counting at one to make log output and settings more readable
while True:
try:
seriesList = _fetchData(pathExpr, startTime, endTime, now, requestContext, seriesList)
break
except Exception:
if retries >= settings.MAX_FETCH_RETRIES:
log.exception("Failed after %s retry! Root cause:\n%s" %
(settings.MAX_FETCH_RETRIES, format_exc()))
raise
else:
log.exception("Got an exception when fetching data! Try: %i of %i. Root cause:\n%s" %
(retries, settings.MAX_FETCH_RETRIES, format_exc()))
retries += 1
return seriesList
def fetch(self, startTime, endTime, now=None, requestContext=None):
try:
data = self.fetch_data(startTime, endTime, now=now)
except IOError:
log.exception("Failed fetch of whisper file '%s'" % self.fs_path)
return None
if not data:
return None
time_info, values = data
(start, end, step) = time_info
meta_info = self.info()
aggregation_method = meta_info['aggregationMethod']
# Merge in data from carbon's cache
values = merge_with_carbonlink(
self.real_metric_path, start, step, values, aggregation_method, self.get_raw_step())
return time_info, values
def index_json(request):
queryParams = request.GET.copy()
queryParams.update(request.POST)
jsonp = queryParams.get('jsonp', False)
cluster = queryParams.get('cluster', False)
def find_matches():
matches = []
for root, dirs, files in os.walk(settings.WHISPER_DIR):
root = root.replace(settings.WHISPER_DIR, '')
for basename in files:
if fnmatch.fnmatch(basename, '*.wsp'):
matches.append(os.path.join(root, basename))
for root, dirs, files in os.walk(settings.CERES_DIR):
root = root.replace(settings.CERES_DIR, '')
for filename in files:
if filename == '.ceres-node':
matches.append(root)
matches = [
m.replace('.wsp', '').replace('.rrd', '').replace('/',
'.').lstrip('.')
for m in sorted(matches)
]
return matches
matches = []
if cluster and len(settings.CLUSTER_SERVERS) >= 1:
try:
matches = reduce( lambda x, y: list(set(x + y)), \
[json.loads(urllib.urlopen('http://' + cluster_server + '/metrics/index.json').read()) \
for cluster_server in settings.CLUSTER_SERVERS])
except urllib.URLError:
log.exception()
return json_response_for(request, matches, jsonp=jsonp, status=500)
else:
matches = find_matches()
return json_response_for(request, matches, jsonp=jsonp)
def auto_complete_values(self,
exprs,
tag,
valuePrefix=None,
limit=None,
requestContext=None):
"""
Return auto-complete suggestions for tags and values based on the matches for the specified expressions, optionally filtered by tag and/or value prefix
"""
if limit is None:
limit = settings.TAGDB_AUTOCOMPLETE_LIMIT
fields = [
('tag', tag or ''),
('valuePrefix', valuePrefix or ''),
('limit', str(limit)),
('local', self.params.get('local', '1')),
]
for expr in exprs:
fields.append(('expr', expr))
result = self.request('/tags/autoComplete/values',
fields,
headers=requestContext.get('forwardHeaders')
if requestContext else None,
timeout=settings.FIND_TIMEOUT)
try:
reader = codecs.getreader('utf-8')
results = json.load(reader(result))
except Exception as err:
self.fail()
log.exception(
"RemoteFinder[%s] Error decoding autocomplete values response from %s: %s"
% (self.host, result.url_full, err))
raise Exception(
"Error decoding autocomplete values response from %s: %s" %
(result.url_full, err))
finally:
result.release_conn()
return results
def merge_with_carbonlink(metric,
start,
step,
values,
aggregation_method=None):
"""Get points from carbonlink and merge them with existing values."""
cached_datapoints = []
try:
cached_datapoints = CarbonLink().query(metric)
except BaseException:
log.exception("Failed CarbonLink query '%s'" % metric)
cached_datapoints = []
if isinstance(cached_datapoints, dict):
cached_datapoints = cached_datapoints.items()
return merge_with_cache(cached_datapoints,
start,
step,
values,
func=aggregation_method)
def _find_paths(self, current_dir, patterns):
"""Recursively generates absolute paths whose components underneath current_dir
match the corresponding pattern in patterns"""
pattern = patterns[0]
patterns = patterns[1:]
try:
entries = os.listdir(current_dir)
except OSError as e:
log.exception(e)
entries = []
subdirs = [e for e in entries if isdir(join(current_dir, e))]
matching_subdirs = match_entries(subdirs, pattern)
if len(
patterns
) == 1 and RRDReader.supported: #the last pattern may apply to RRD data sources
files = [e for e in entries if isfile(join(current_dir, e))]
rrd_files = match_entries(files, pattern + ".rrd")
if rrd_files: #let's assume it does
datasource_pattern = patterns[0]
for rrd_file in rrd_files:
absolute_path = join(current_dir, rrd_file)
yield absolute_path + self.DATASOURCE_DELIMETER + datasource_pattern
if patterns: #we've still got more directories to traverse
for subdir in matching_subdirs:
absolute_path = join(current_dir, subdir)
for match in self._find_paths(absolute_path, patterns):
yield match
else: #we've got the last pattern
files = [e for e in entries if isfile(join(current_dir, e))]
matching_files = match_entries(files, pattern + '.*')
for basename in matching_files + matching_subdirs:
yield join(current_dir, basename)
def get_results(self):
if self.failed:
return
if self.cachedResult is not None:
results = self.cachedResult
else:
if self.connection is None:
self.send()
try:
try: # Python 2.7+, use buffering of HTTP responses
response = self.connection.getresponse(buffering=True)
except TypeError: # Python 2.6 and older
response = self.connection.getresponse()
assert response.status == 200, "received error response %s - %s" % (
response.status, response.reason)
result_data = response.read()
results = unpickle.loads(result_data)
except:
log.exception(
"FindRequest.get_results(host=%s, query=%s) exception processing response"
% (self.store.host, self.query))
self.store.fail()
return
cache.set(self.cacheKey, results, settings.FIND_CACHE_DURATION)
for node_info in results:
if node_info.get('is_leaf'):
reader = RemoteReader(self.store,
node_info,
bulk_query=self.query.pattern)
node = LeafNode(node_info['path'], reader)
else:
node = BranchNode(node_info['path'])
node.local = False
yield node
def accessor(self):
"""Return an accessor."""
with self._lock:
if not self._accessor:
from django.conf import settings as django_settings
accessor = graphite_utils.accessor_from_settings(
django_settings)
# If connect() fail it will raise an exception that will be caught
# by the caller. If the plugin is called again, self._accessor will
# still be None and a new accessor will be created.
try:
accessor.connect()
except Exception as e:
log.exception("failed to connect()")
accessor.shutdown()
raise e
accessor.set_cache(
bg_accessor_cache.DjangoCache(self.django_cache()),
metadata_ttl=django_settings.FIND_CACHE_DURATION,
data_ttl=django_settings.DEFAULT_CACHE_DURATION)
self._accessor = accessor
return self._accessor
def send(self, headers=None):
log.info("FindRequest.send(host=%s, query=%s) called" %
(self.store.host, self.query))
if headers is None:
headers = {}
self.cachedResult = cache.get(self.cacheKey)
if self.cachedResult is not None:
log.info("FindRequest(host=%s, query=%s) using cached result" %
(self.store.host, self.query))
return
query_params = [
('local', '1'),
('format', 'pickle'),
('query', self.query.pattern),
]
if self.query.startTime:
query_params.append(('from', self.query.startTime))
if self.query.endTime:
query_params.append(('until', self.query.endTime))
query_string = urlencode(query_params)
try:
connector_class = connector_class_selector(
settings.INTRACLUSTER_HTTPS)
self.connection = connector_class(self.store.host)
self.connection.timeout = settings.REMOTE_FIND_TIMEOUT
self.connection.request('GET', '/metrics/find/?' + query_string,
None, headers)
except:
log.exception(
"FindRequest.send(host=%s, query=%s) exception during request"
% (self.store.host, self.query))
self.store.fail()
self.failed = True
def prefetchRemoteData(requestContext, pathExpressions):
# if required, fetch data from all remote nodes
# storing the result in a big hash of the form:
# data[node][hash(originalPathExpression, start, end)] = [ matchingSeries, matchingSeries2, ... ]
prefetchedRemoteData = {}
if requestContext['localOnly']:
return prefetchedRemoteData
(startTime, endTime, now) = _timebounds(requestContext)
result_queue = fetchRemoteData(requestContext, pathExpressions, False)
while not result_queue.empty():
try:
(node, results) = result_queue.get_nowait()
except Queue.Empty:
log.exception("result_queue not empty, but unable to retrieve results")
# prefill result with empty list
# Needed to be able to detect if a query has already been made
prefetchedRemoteData[node] = {}
for pe in pathExpressions:
prefetchedRemoteData[node][_prefetchMetricKey(pe, startTime, endTime)] = []
for series in results:
# series.pathExpression is original target, ie. containing wildcards
# XXX would be nice to disable further prefetch calls to that backend
try:
k = _prefetchMetricKey(series['pathExpression'], startTime, endTime)
except KeyError:
log.exception("Remote node %s doesn't support prefetching data... upgrade!" % node)
raise
if prefetchedRemoteData[node].get(k) is None:
# This should not be needed because of above filling with [],
# but could happen if backend sends unexpected stuff
prefetchedRemoteData[node][k] = [series]
else:
prefetchedRemoteData[node][k].append(series)
return prefetchedRemoteData
def set_metadata_view(request):
results = {}
if request.method == 'GET':
metric = request.GET['metric']
key = request.GET['key']
value = request.GET['value']
try:
results[metric] = CarbonLink.set_metadata(metric, key, value)
except:
log.exception()
results[metric] = dict(
error=
"Unexpected error occurred in CarbonLink.set_metadata(%s, %s)"
% (metric, key))
elif request.method == 'POST':
if request.META.get('CONTENT_TYPE') == 'application/json':
operations = json.loads(request.body)
else:
operations = json.loads(request.POST['operations'])
for op in operations:
metric = None
try:
metric, key, value = op['metric'], op['key'], op['value']
results[metric] = CarbonLink.set_metadata(metric, key, value)
except:
log.exception()
if metric:
results[metric] = dict(
error=
"Unexpected error occurred in bulk CarbonLink.set_metadata(%s)"
% metric)
else:
results = dict(error='Invalid request method')
return json_response_for(request, results)
def fetch(self, startTime, endTime):
try:
data = self.fetch_data(startTime, endTime)
except IOError:
log.exception("Failed fetch of whisper file '%s'" % self.fs_path)
return None
if not data:
return None
time_info, values = data
(start, end, step) = time_info
meta_info = self.info()
aggregation_method = meta_info['aggregationMethod']
lowest_step = min([i['secondsPerPoint']
for i in meta_info['archives']])
# Merge in data from carbon's cache
values = merge_with_carbonlink(
self.real_metric_path, start, step, values, aggregation_method)
return time_info, values
def fetch(self, startTime, endTime):
data = self.ceres_node.read(startTime, endTime)
time_info = (data.startTime, data.endTime, data.timeStep)
values = list(data.values)
# Merge in data from carbon's cache
try:
cached_datapoints = CarbonLink.query(self.real_metric_path)
except:
log.exception("Failed CarbonLink query '%s'" % self.real_metric_path)
cached_datapoints = []
for (timestamp, value) in cached_datapoints:
interval = timestamp - (timestamp % data.timeStep)
try:
i = int(interval - data.startTime) / data.timeStep
values[i] = value
except:
pass
return (time_info, values)
def wait_for_results():
if request_lock.acquire(False): # we only send the request the first time we're called
try:
log.info("RemoteReader.request_data :: requesting %s" % url)
self.connection = HTTPConnectionWithTimeout(self.store.host)
self.connection.timeout = settings.REMOTE_FETCH_TIMEOUT
self.connection.request('GET', urlpath)
except:
log.exception("Error requesting %s" % url)
wait_lock.acquire(False)
completion_event.set()
self.store.fail()
raise
if wait_lock.acquire(False): # the FetchInProgress that gets waited on waits for the actual completion
try:
response = self.connection.getresponse()
if response.status != 200:
raise Exception("Error response %d %s from %s" % (response.status, response.reason, url))
pickled_response = response.read()
results = unpickle.loads(pickled_response)
self.cache_lock.acquire()
self.request_cache[url] = results
self.cache_lock.release()
completion_event.set()
return results
except:
completion_event.set()
self.store.fail()
log.exception("Error requesting %s" % url)
raise
else: # otherwise we just wait on the completion_event
completion_event.wait(settings.REMOTE_FETCH_TIMEOUT)
cached_results = self.request_cache.get(url)
if cached_results is None:
raise Exception("Passive remote fetch failed to find cached results")
else:
return cached_results
def receiveResponse():
try:
buf = ''
remaining = 4
message_size = None
while remaining:
packet = connection.recv(remaining)
assert packet, "CarbonLink lost connection to %s" % str(
host)
buf += packet
if message_size is None:
if len(buf) == 4:
remaining = message_size = struct.unpack(
"!L", buf)[0]
buf = ''
continue
remaining -= len(packet)
# We're done with the connection for this request, put it in the pool
self.putConnectionInPool(host, connection)
# Now parse the response
points = pickle.loads(buf)
log.cache("CarbonLink to %s, retrieved %d points for %s" %
(host, len(points), metric))
for point in points:
yield point
except:
log.exception(
"CarbonLink to %s, exception while getting response" %
str(host))
self.removeConnectionFromPool(host, connection)
def fetch_multi(self, startTime, endTime, now=None, requestContext=None):
if not self.bulk_query:
return []
query_params = [('format', self.finder.params.get('format', 'pickle')),
('local', self.finder.params.get('local', '1')),
('noCache', '1'), ('from', int(startTime)),
('until', int(endTime))]
for target in self.bulk_query:
query_params.append(('target', target))
if now is not None:
query_params.append(('now', int(now)))
headers = requestContext.get(
'forwardHeaders') if requestContext else None
retries = 1 # start counting at one to make log output and settings more readable
while True:
try:
result = self.finder.request(
'/render/',
fields=query_params,
headers=headers,
timeout=settings.FETCH_TIMEOUT,
)
break
except Exception:
if retries >= settings.MAX_FETCH_RETRIES:
log.exception("Failed after %s attempts! Root cause:\n%s" %
(settings.MAX_FETCH_RETRIES, format_exc()))
raise
else:
log.exception(
"Got an exception when fetching data! Try: %i of %i. Root cause:\n%s"
% (retries, settings.MAX_FETCH_RETRIES, format_exc()))
retries += 1
data = self.deserialize(result)
try:
return [{
'pathExpression':
series.get('pathExpression', series['name']),
'name':
series['name'],
'time_info': (series['start'], series['end'], series['step']),
'values':
series['values'],
} for series in data]
except Exception as err:
self.finder.fail()
log.exception(
"RemoteReader[%s] Invalid render response from %s: %s" %
(self.finder.host, result.url_full, repr(err)))
raise Exception("Invalid render response from %s: %s" %
(result.url_full, repr(err)))
def wait_for_results():
connection_event.wait(1)
connection = self.connections.get(url)
if not connection:
log.exception("Wait For Results - No connection found.")
if wait_lock.acquire(
False
): # the FetchInProgress that gets waited on waits for the actual completion
try:
response = connection.getresponse()
if response.status != 200:
raise Exception(
"Error response %d %s from %s" %
(response.status, response.reason, url))
pickled_response = response.read()
results = pickle.loads(pickled_response)
self.cache_lock.acquire()
self.request_cache[url] = results
self.cache_lock.release()
completion_event.set()
return results
except:
completion_event.set()
self.store.fail()
log.exception("Wait For Results - Error requesting %s" %
url)
raise
else: # otherwise we just wait on the completion_event
completion_event.wait(settings.REMOTE_FETCH_TIMEOUT)
cached_results = self.request_cache.get(url)
if cached_results is None:
raise Exception(
"Passive remote fetch failed to find cached results for %s"
% url)
else:
return cached_results
def fetch(self, startTime, endTime):
data = whisper.fetch(self.fs_path, startTime, endTime)
time_info, values = data
(start, end, step) = time_info
# Merge in data from carbon's cache
try:
cached_datapoints = CarbonLink.query(self.real_metric_path)
except:
log.exception("Failed CarbonLink query '%s'" %
self.real_metric_path)
cached_datapoints = []
for (timestamp, value) in cached_datapoints:
interval = timestamp - (timestamp % step)
try:
i = int(interval - start) / step
values[i] = value
except:
pass
return (time_info, values)
def get_index(self, requestContext):
url = '/metrics/index.json'
headers = requestContext.get('forwardHeaders')
result = self.request(
url,
fields=[
('local', '1'),
],
headers=headers,
timeout=settings.REMOTE_FIND_TIMEOUT)
try:
results = json.loads(result.data)
except Exception as err:
self.fail()
log.exception(
"RemoteFinder[%s] Error decoding index response from %s: %s" %
(self.host, result.url_full, err))
raise Exception("Error decoding index response from %s: %s" % (result.url_full, err))
return results
