Rush has uploaded a new change for review.
https://gerrit.wikimedia.org/r/239511
Change subject: WIP: elastic: sane diamond collector
......................................................................
WIP: elastic: sane diamond collector
ref T111573
Change-Id: I9cefed6ee1ac7cbcf39a28e611bbe28bcff402f4
---
A modules/elasticsearch/files/monitor/wmfelastic.py
A modules/elasticsearch/manifests/monitor/diamond.pp
2 files changed, 218 insertions(+), 0 deletions(-)
git pull ssh://gerrit.wikimedia.org:29418/operations/puppet
refs/changes/11/239511/1
diff --git a/modules/elasticsearch/files/monitor/wmfelastic.py
b/modules/elasticsearch/files/monitor/wmfelastic.py
new file mode 100644
index 0000000..bbb7153
--- /dev/null
+++ b/modules/elasticsearch/files/monitor/wmfelastic.py
@@ -0,0 +1,209 @@
+# coding=utf-8
+
+"""
+2015 Chase Pettet
+
+This is a significantly paired down
+collector for the WMF scale. The native
+diamond collector has dozens of stats per
+index which at our scale is impractical at
+the moment.
+
+This collector publishes a set list of metrics
+and only publishes cluster metrics from the
+current master.
+
+"""
+import urllib2
+
+try:
+ import json
+ json # workaround for pyflakes issue #13
+except ImportError:
+ import simplejson as json
+
+import diamond.collector
+
+
+class WMFElasticCollector(diamond.collector.Collector):
+
+ def __init__(self, *args, **kwargs):
+ super(WMFElasticCollector, self).__init__(*args, **kwargs)
+
+ self.node_id = ''
+
+ self.endpoints = {
+ 'node': '_nodes/_local/stats',
+ 'cluster': '_cluster/stats',
+ }
+
+ # Metrics provided at cluster and node level
+ # _cluster/stats
+ self.cluster_metrics = [
+ 'indices.docs.deleted',
+ 'indices.docs.count',
+ 'indices.count',
+ 'indices.shards.total',
+ 'indices.shards.primaries',
+ 'indices.shards.replication',
+ 'indices.store.size_in_bytes',
+ 'indices.segments.count',
+ 'indices.segments.memory_in_bytes',
+ 'indices.segments.index_writer_max_memory_in_bytes',
+ 'indices.segments.version_map_memory_in_bytes',
+ 'indices.segments.fixed_bit_set_memory_in_bytes',
+ 'nodes.count.total',
+ 'nodes.os.mem.total_in_bytes',
+ 'nodes.jvm.mem.heap_used_in_bytes',
+ 'nodes.jvm.mem.heap_max_in_bytes',
+ 'nodes.jvm.threads',
+ 'nodes.fs.total_in_bytes',
+ 'nodes.fs.free_in_bytes',
+ 'nodes.fs.available_in_bytes',
+ 'nodes.fs.disk_reads',
+ 'nodes.fs.disk_writes',
+ 'nodes.fs.disk_io_op',
+ ]
+
+ # node level metrics
+ # '_nodes/_local/stats'
+ self.node_metrics = [
+ 'indices.docs.deleted',
+ 'indices.docs.count',
+ 'indices.store.size_in_bytes',
+ 'indices.store.throttle_time_in_millis',
+ 'indices.indexing.index_total',
+ 'indices.indexing.index_time_in_millis',
+ 'indices.indexing.index_current',
+ 'indices.indexing.delete_total',
+ 'indices.indexing.delete_time_in_millis',
+ 'indices.indexing.delete_current',
+ 'indices.indexing.throttle_time_in_millis',
+ 'indices.get.total',
+ 'indices.get.time_in_millis',
+ 'indices.get.exists_total',
+ 'indices.get.missing_total',
+ 'indices.get.missing_time_in_millis',
+ 'indices.get.current',
+ 'indices.search.open_contexts',
+ 'indices.search.query_total',
+ 'indices.search.query_time_in_millis',
+ 'indices.search.query_current',
+ 'indices.search.fetch_total',
+ 'indices.merges.current',
+ 'indices.merges.current_docs',
+ 'indices.merges.current_size_in_bytes',
+ 'indices.merges.total',
+ 'indices.merges.total_time_in_millis',
+ 'indices.merges.total_docs',
+ 'indices.merges.total_size_in_bytes',
+ 'indices.refresh.total',
+ 'indices.refresh.total_time_in_millis',
+ 'indices.flush.total',
+ 'indices.flush.total_time_in_millis',
+ 'indices.warmer.current',
+ 'indices.warmer.total',
+ 'indices.warmer.total_time_in_millis',
+
+ 'process.open_file_descriptors',
+
+ 'jvm.mem.heap_used_in_bytes',
+ 'jvm.mem.heap_used_percent',
+ 'jvm.mem.heap_committed_in_bytes',
+ 'jvm.mem.heap_max_in_bytes',
+ 'jvm.mem.non_heap_used_in_bytes',
+ 'jvm.mem.non_heap_committed_in_bytes',
+ 'jvm.threads.count',
+ 'jvm.threads.peak_count',
+ 'jvm.gc.collectors.young.collection_count',
+ 'jvm.gc.collectors.young.collection_time_in_millis',
+ 'jvm.gc.collectors.old.collection_count',
+ 'jvm.gc.collectors.old.collection_time_in_millis',
+
+ 'http.current_open',
+ 'http.total_opened',
+ ]
+
+ def get_default_config_help(self):
+ config_help = super(WMFElasticCollector,
+ self).get_default_config_help()
+ config_help.update({
+ 'host': 'localhost',
+ 'port': '9200',
+ })
+ return config_help
+
+ def get_default_config(self):
+ """
+ Returns the default collector settings
+ """
+ config = super(WMFElasticCollector, self).get_default_config()
+ config.update({
+ 'host': '127.0.0.1',
+ 'port': 9200,
+ 'path': 'elasticsearch',
+ 'stats': ['node', 'cluster'],
+ })
+ return config
+
+ def _get(self, path):
+ url = 'http://%s:%i/%s' % (self.config['host'],
+ self.config['port'],
+ path)
+ response = urllib2.urlopen(url)
+ return json.load(response)
+
+ def is_master(self):
+ master = self._get('_cluster/state/master_node')
+ if self.node_id == master['master_node']:
+ return master['cluster_name']
+
+ def dict_digger(self, depth, ddict):
+ if len(depth) == 1:
+ return ddict[depth[0]]
+ else:
+ key = depth.pop(0)
+ return self.dict_digger(depth, ddict[key])
+
+ def cluster_stats(self):
+ cstats = self._get(self.endpoints['cluster'])
+ gmetrics = {}
+ for metric in self.cluster_metrics:
+ depth = metric.split('.')
+ try:
+ value = self.dict_digger(depth, cstats)
+ gmetrics[metric] = value
+ except KeyError, e:
+ pass
+ return gmetrics
+
+ def node_stats(self):
+ astats = self._get(self.endpoints['node'])
+ self.node_id = astats['nodes'].keys()[0]
+ nodestats = astats['nodes'][self.node_id]
+ gmetrics = {}
+ for metric in self.node_metrics:
+ depth = metric.split('.')
+ try:
+ value = self.dict_digger(depth, nodestats)
+ gmetrics[metric] = value
+ except KeyError, e:
+ pass
+ return gmetrics
+
+ def collect(self):
+ node_stats = self.node_stats()
+ for metric, value in node_stats.iteritems():
+ self.publish(metric, value)
+
+ master = self.is_master()
+ if not master:
+ return
+
+ # Set the prepend to override "servers"
+ self.config['path_prefix'] = 'clusters'
+ # Set the hostname to the name of the cluster
+ self.config['hostname'] = master
+ cluster_stats = self.cluster_stats()
+ for metric, value in cluster_stats.iteritems():
+ self.publish(metric, value)
diff --git a/modules/elasticsearch/manifests/monitor/diamond.pp
b/modules/elasticsearch/manifests/monitor/diamond.pp
new file mode 100644
index 0000000..86b2884
--- /dev/null
+++ b/modules/elasticsearch/manifests/monitor/diamond.pp
@@ -0,0 +1,9 @@
+# == Class: elasticsearch::monitor::diamond
+#
+
+class elasticsearch::monitor::diamond {
+ diamond::collector { 'WMFElastic':
+ source => 'puppet:///modules/elasticsearch/monitor/wmfelastic.py',
+ }
+}
+
--
To view, visit https://gerrit.wikimedia.org/r/239511
To unsubscribe, visit https://gerrit.wikimedia.org/r/settings
Gerrit-MessageType: newchange
Gerrit-Change-Id: I9cefed6ee1ac7cbcf39a28e611bbe28bcff402f4
Gerrit-PatchSet: 1
Gerrit-Project: operations/puppet
Gerrit-Branch: production
Gerrit-Owner: Rush <[email protected]>
_______________________________________________
MediaWiki-commits mailing list
[email protected]
https://lists.wikimedia.org/mailman/listinfo/mediawiki-commits
