Skip to content

Instantly share code, notes, and snippets.

@anatolijd
Last active December 16, 2017 04:48
Show Gist options
  • Save anatolijd/5611927 to your computer and use it in GitHub Desktop.
Save anatolijd/5611927 to your computer and use it in GitHub Desktop.
Following the https://gist.github.com/anatolijd/5576536 and https://answers.launchpad.net/graphite/+question/228472 I've limited the combinations (each CARBONLINK_HOSTS item is assigned a list of local carbon-cache instances only), and rearranged the output. Now it is pretty clear visible that webapp consistent hashing get_node(key) does the job…
# python hash_ring_selector2.py
DESTINATIONS : ['10.4.0.1:2104:a', '10.4.0.1:2204:b',
'10.4.0.2:2104:a', '10.4.0.2:2204:b']
CARBONLINK_HOSTS: ['10.4.0.1:7102:a', '10.4.0.1:7202:b']
result metric carbon-relay webapp
[ OK ] some.metric ('10.4.0.1', 'b') ('10.4.0.1', 'b')
[ OK ] mymetric ('10.4.0.1', 'a') ('10.4.0.1', 'a')
[ OK ] test.metric2 ('10.4.0.1', 'b') ('10.4.0.1', 'b')
[ OK ] test.metric3 ('10.4.0.1', 'b') ('10.4.0.1', 'b')
[ OK ] test.metric4 ('10.4.0.1', 'b') ('10.4.0.1', 'b')
[ OK ] test.tolik ('10.4.0.1', 'a') ('10.4.0.1', 'a')
[ OK ] test.natasha.z5poverhu.dura ('10.4.0.1', 'a') ('10.4.0.1', 'a')
[ OK ] system.loadavg_5min ('10.4.0.1', 'a') ('10.4.0.1', 'a')
[ OK ] system.loadavg_1min ('10.4.0.1', 'b') ('10.4.0.1', 'b')
[ OK ] carbon.relays.graphite-1-a.memUsage ('10.4.0.1', 'a') ('10.4.0.1', 'a')
[ OK ] carbon.relays.graphite-1-a.metricsReceived ('10.4.0.1', 'b') ('10.4.0.1', 'b')
[ OK ] secondYAxis(carbon.agents.*.pointsPerUpdate) ('10.4.0.1', 'b') ('10.4.0.1', 'b')
[ OK ] sumSeries(group(carbon.agents.*.creates)) ('10.4.0.1', 'a') ('10.4.0.1', 'a')
[ OK ] net.iostat.mongo-rfcalaiscache.rks ('10.4.0.1', 'a') ('10.4.0.1', 'a')
CARBONLINK_HOSTS: ['10.4.0.2:7102:a', '10.4.0.2:7202:b']
result metric carbon-relay webapp
[ OK ] test.metric ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] test.metric1 ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] test.anatoliy ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] test.random ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] test.roma ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] test.slavik ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] test.valera ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] system.loadavg_15min ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] carbon.relays.graphite-1-a.cpuUsage ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] carbon.agents.graphite-1-b.memUsage ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] carbon.agents.graphite-1-b.cpuUsage ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] carbon.agents.graphite-1-b.metricsReceived ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] group(carbon.agents.*.metricsReceived) ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] stats_cos.ha-69b.ranker.source_cache.load_att ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] chef.es-66b.updated_resources ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] disk.es-63b._opt_elasticsearch_data ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] chef.es-*.updated_resources ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] disk.es-63b._opt_elasticsearch_data.used ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] hosts.es-67b.disk-xvdc.disk_ops.read ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] hosts.es-67b.procs.logstash_agent_3.rss ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] hosts.mon-80b.procs.backup_mon_8220b.cputime ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] www.jsp.anonymous.act.ClickEvent.1MinuteRate ('10.4.0.2', 'a') ('10.4.0.2', 'a')
[ OK ] www.jsp.external.act.ResultsAutoAnytime.count ('10.4.0.2', 'b') ('10.4.0.2', 'b')
# python hash_ring_selector2.py
DESTINATIONS : ['10.4.0.1:2104:a', '10.4.0.1:2204:b', '10.4.0.1:2304:c',
'10.4.0.2:2104:a', '10.4.0.2:2204:b', '10.4.0.2:2304:c',
'10.4.0.3:2104:a', '10.4.0.3:2204:b', '10.4.0.3:2304:c']
CARBONLINK_HOSTS: ['10.4.0.1:7102:a', '10.4.0.1:7202:b', '10.4.0.1:7302:c']
result metric carbon-relay webapp
[ OK ] some.metric ('10.4.0.1', 'c') ('10.4.0.1', 'c')
[ OK ] mymetric ('10.4.0.1', 'a') ('10.4.0.1', 'a')
[ OK ] test.valera ('10.4.0.1', 'c') ('10.4.0.1', 'c')
[ OK ] test.natasha.z5poverhu.dura ('10.4.0.1', 'a') ('10.4.0.1', 'a')
[ OK ] system.loadavg_5min ('10.4.0.1', 'a') ('10.4.0.1', 'a')
[ OK ] carbon.agents.graphite-1-b.metricsReceived ('10.4.0.1', 'c') ('10.4.0.1', 'c')
[ OK ] secondYAxis(carbon.agents.*.pointsPerUpdate) ('10.4.0.1', 'b') ('10.4.0.1', 'b')
[ OK ] sumSeries(group(carbon.agents.*.creates)) ('10.4.0.1', 'a') ('10.4.0.1', 'a')
CARBONLINK_HOSTS: ['10.4.0.2:7102:a', '10.4.0.2:7202:b', '10.4.0.2:7302:c']
result metric carbon-relay webapp
[ OK ] test.metric1 ('10.4.0.2', 'c') ('10.4.0.2', 'c')
[ OK ] test.metric4 ('10.4.0.2', 'c') ('10.4.0.2', 'c')
[ OK ] test.anatoliy ('10.4.0.2', 'c') ('10.4.0.2', 'c')
[ OK ] test.random ('10.4.0.2', 'c') ('10.4.0.2', 'c')
[ OK ] system.loadavg_1min ('10.4.0.2', 'c') ('10.4.0.2', 'c')
[ OK ] carbon.agents.graphite-1-b.memUsage ('10.4.0.2', 'c') ('10.4.0.2', 'c')
[ OK ] chef.es-66b.updated_resources ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] disk.es-63b._opt_elasticsearch_data ('10.4.0.2', 'c') ('10.4.0.2', 'c')
[ OK ] chef.es-*.updated_resources ('10.4.0.2', 'c') ('10.4.0.2', 'c')
[ OK ] disk.es-63b._opt_elasticsearch_data.used ('10.4.0.2', 'c') ('10.4.0.2', 'c')
[ OK ] hosts.es-67b.disk-xvdc.disk_ops.read ('10.4.0.2', 'b') ('10.4.0.2', 'b')
[ OK ] www.jsp.external.act.ResultsAutoAnytime.count ('10.4.0.2', 'c') ('10.4.0.2', 'c')
CARBONLINK_HOSTS: ['10.4.0.3:7102:a', '10.4.0.3:7202:b', '10.4.0.3:7302:c']
result metric carbon-relay webapp
[ OK ] test.metric ('10.4.0.3', 'b') ('10.4.0.3', 'b')
[ OK ] test.metric2 ('10.4.0.3', 'b') ('10.4.0.3', 'b')
[ OK ] test.metric3 ('10.4.0.3', 'a') ('10.4.0.3', 'a')
[ OK ] test.roma ('10.4.0.3', 'c') ('10.4.0.3', 'c')
[ OK ] test.slavik ('10.4.0.3', 'a') ('10.4.0.3', 'a')
[ OK ] test.tolik ('10.4.0.3', 'c') ('10.4.0.3', 'c')
[ OK ] system.loadavg_15min ('10.4.0.3', 'a') ('10.4.0.3', 'a')
[ OK ] carbon.relays.graphite-1-a.memUsage ('10.4.0.3', 'a') ('10.4.0.3', 'a')
[ OK ] carbon.relays.graphite-1-a.cpuUsage ('10.4.0.3', 'a') ('10.4.0.3', 'a')
[ OK ] carbon.relays.graphite-1-a.metricsReceived ('10.4.0.3', 'a') ('10.4.0.3', 'a')
[ OK ] carbon.agents.graphite-1-b.cpuUsage ('10.4.0.3', 'b') ('10.4.0.3', 'b')
[ OK ] group(carbon.agents.*.metricsReceived) ('10.4.0.3', 'c') ('10.4.0.3', 'c')
[ OK ] net.iostat.mongo-rfcalaiscache.rks ('10.4.0.3', 'b') ('10.4.0.3', 'b')
[ OK ] stats_cos.ha-69b.ranker.source_cache.load_att ('10.4.0.3', 'c') ('10.4.0.3', 'c')
[ OK ] hosts.es-67b.procs.logstash_agent_3.rss ('10.4.0.3', 'b') ('10.4.0.3', 'b')
[ OK ] hosts.mon-80b.procs.backup_mon_8220b.cputime ('10.4.0.3', 'a') ('10.4.0.3', 'a')
[ OK ] www.jsp.anonymous.act.ClickEvent.1MinuteRate ('10.4.0.3', 'a') ('10.4.0.3', 'a')
#!/bin/python
import os,sys
sys.path.append('/opt/graphite/webapp')
os.environ['DJANGO_SETTINGS_MODULE'] = 'graphite.settings'
from graphite.storage import STORE, LOCAL_STORE
from graphite.render.hashing import ConsistentHashRing
from graphite.render.datalib import CarbonLinkPool
# Use CarbonLinkPool class, as it already contains the same routing methods that are used by either carbon-relay or webapp
def build_hosts(hosts=[]):
res = []
for host in hosts:
parts = host.split(':')
server = parts[0]
port = int( parts[1] )
if len(parts) > 2:
instance = parts[2]
else:
instance = None
res.append( (server, int(port), instance) )
return res
def compare_routes(WRITE,READ,metrics=[],verbose=True):
print " result %45s %20s %20s" % ('metric','carbon-relay','webapp')
status = True
for metric in metrics:
wi = WRITE.select_host(metric)
ri = READ.select_host(metric)
res = " [ %4s ] %45s %20s %20s" % ('OK' if (wi == ri) else 'Fail', metric, wi, ri)
(wh,wc) = wi ; (rh,rc) = ri # compare only matching hosts
if wh != rh:
continue
if wi == ri:
if verbose:
print res
else:
# print only failed matches
status = False
print res
return status
CARBONLINK_TIMEOUT = 60
# CARBONLINK_HOSTS for 2 nodes
CARBONLINK_HOSTS = [
['10.4.0.1:7102:a','10.4.0.1:7202:b','10.4.0.1:7302:c'], # 1-st node in a cluster
['10.4.0.2:7102:a','10.4.0.2:7202:b','10.4.0.2:7302:c'], # 2nd node in a cluster
['10.4.0.3:7102:a','10.4.0.3:7202:b','10.4.0.3:7302:c'], # 3rd node in a cluster
]
# all carbon-cache instances
DESTINATIONS = [
['10.4.0.1:2104:a','10.4.0.1:2204:b','10.4.0.1:2304:c',
'10.4.0.2:2104:a','10.4.0.2:2204:b','10.4.0.2:2304:c',
'10.4.0.3:2104:a','10.4.0.3:2204:b','10.4.0.3:2304:c']
# ['10.4.0.1:2104:a','10.4.0.1:2204:b','10.4.0.2:2104:a','10.4.0.2:2204:b']
]
# metrics to test
metrics = [
"some.metric",
"mymetric",
"test.metric",
"test.metric1",
"test.metric2",
"test.metric3",
"test.metric4",
"test.anatoliy",
"test.random",
"test.roma",
"test.slavik",
"test.tolik",
"test.valera",
"test.natasha.z5poverhu.dura",
"system.loadavg_15min",
"system.loadavg_5min",
"system.loadavg_1min",
"carbon.relays.graphite-1-a.memUsage",
"carbon.relays.graphite-1-a.cpuUsage",
"carbon.relays.graphite-1-a.metricsReceived",
"carbon.agents.graphite-1-b.memUsage",
"carbon.agents.graphite-1-b.cpuUsage",
"carbon.agents.graphite-1-b.metricsReceived",
"group(carbon.agents.*.metricsReceived)",
"secondYAxis(carbon.agents.*.pointsPerUpdate)",
"sumSeries(group(carbon.agents.*.creates))",
"net.iostat.mongo-rfcalaiscache.rks",
"stats_cos.ha-69b.ranker.source_cache.load_att",
"chef.es-66b.updated_resources",
"disk.es-63b._opt_elasticsearch_data",
"chef.es-*.updated_resources",
"disk.es-63b._opt_elasticsearch_data.used",
"hosts.es-67b.disk-xvdc.disk_ops.read",
"hosts.es-67b.procs.logstash_agent_3.rss",
"hosts.mon-80b.procs.backup_mon_8220b.cputime",
"www.jsp.anonymous.act.ClickEvent.1MinuteRate",
"www.jsp.external.act.ResultsAutoAnytime.count"
]
# For every metric, emulate ConsistentHashRing.select_host(metric) for carbon-relay (WRITE) and webapp (READ).
# Use many different metrics for better representative.
WRITE = CarbonLinkPool(build_hosts(DESTINATIONS[0]), CARBONLINK_TIMEOUT)
print "\nDESTINATIONS : %s" % DESTINATIONS[0]
# webapp query carbon-cache at 1-st node
READ = CarbonLinkPool(build_hosts(CARBONLINK_HOSTS[0]), CARBONLINK_TIMEOUT)
print "\nCARBONLINK_HOSTS: %s\n" % CARBONLINK_HOSTS[0]
compare_routes(WRITE,READ,metrics,True)
# webapp query cache at 2-nd node
READ = CarbonLinkPool(build_hosts(CARBONLINK_HOSTS[1]), CARBONLINK_TIMEOUT)
print "\nCARBONLINK_HOSTS: %s\n" % CARBONLINK_HOSTS[1]
compare_routes(WRITE,READ,metrics,True)
# webapp query cache at 3-rd node
READ = CarbonLinkPool(build_hosts(CARBONLINK_HOSTS[2]), CARBONLINK_TIMEOUT)
print "\nCARBONLINK_HOSTS: %s\n" % CARBONLINK_HOSTS[2]
compare_routes(WRITE,READ,metrics,True)
@marco-hoyer
Copy link

Hi,

thanks a lot for your work. May it be that there was an older version of carbon, used for this script?
I used carbon/master and faced an import error for CarbonLinkPool. Following import statement works for me:

from graphite.carbonlink import CarbonLinkPool

greets

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment