From Fedora Project Wiki
No edit summary |
No edit summary |
||
| Line 45: | Line 45: | ||
Then apply this to all cpu_util counters, saving the result in the 'average_util' collection: | Then apply this to all cpu_util counters, saving the result in the 'average_util' collection: | ||
<nowiki> | <nowiki> | ||
> db.meter.mapReduce(mapVolumes, reduceAverage, { out: { merge: 'average_util' }, query: {'counter_name': 'cpu_util'}}); | > db.meter.mapReduce(mapVolumes, reduceAverage, { out: { merge: 'average_util' }, query: {'counter_name': 'cpu_util'}}); | ||
{ | { | ||
"result" : " | "result" : "average_util", | ||
"timeMillis" : 110, | "timeMillis" : 110, | ||
"counts" : { | "counts" : { | ||
| Line 58: | Line 58: | ||
"ok" : 1, | "ok" : 1, | ||
} | } | ||
</nowiki> | </nowiki> | ||
Display this temporary collection: | Display this temporary collection: | ||
Revision as of 15:54, 2 April 2013
Ceilometer uses mongoDB by default to store metering data (though storage drivers are also supported for sqlalchemy and more recently for HBase too).
In this test we use the mongo client to explore the metering store:
mongo ceilometer
Show the available collections:
> show collections meter project resource system.indexes user
Note the default indices:
> db.system.indexes.find()
{ "v" : 1, "key" : { "_id" : 1 }, "ns" : "ceilometer.resource", "name" : "_id_" }
{ "v" : 1, "key" : { "user_id" : 1, "source" : 1 }, "ns" : "ceilometer.resource", "name" : "resource_idx" }
{ "v" : 1, "key" : { "_id" : 1 }, "ns" : "ceilometer.meter", "name" : "_id_" }
{ "v" : 1, "key" : { "resource_id" : 1, "user_id" : 1, "counter_name" : 1, "timestamp" : 1, "source" : 1 },
"ns" : "ceilometer.meter", "name" : "meter_idx" }
{ "v" : 1, "key" : { "_id" : 1 }, "ns" : "ceilometer.user", "name" : "_id_" }
{ "v" : 1, "key" : { "_id" : 1 }, "ns" : "ceilometer.project", "name" : "_id_" }
Show the cpu_util samples retrived earlier via the ceilometer CLI, sorted in ascending time-order:
> db.meter.find({'counter_name': 'cpu_util', 'resource_id' : $INSTANCE_ID}).sort({'timestamp': 1})
{ "_id" : ObjectId("515af62a1a63121d32eaa00c"), "counter_name" : "cpu_util", "user_id" : ...
Do some simple map-reduce to determine the average CPU utilization across all instances. First define javascript functions to map the counter volumes and calculate the average respectively:
> var mapVolumes = function() {
emit(this.resource_id, this.counter_volume);
}
> var reduceAverage = function (key, values) {
var total = 0;
for (var i = 0; i < values.length; i++) {
total += values[i];
}
return total / values.length;
}
Then apply this to all cpu_util counters, saving the result in the 'average_util' collection:
> db.meter.mapReduce(mapVolumes, reduceAverage, { out: { merge: 'average_util' }, query: {'counter_name': 'cpu_util'}});
{
"result" : "average_util",
"timeMillis" : 110,
"counts" : {
"input" : 22,
"emit" : 22,
"reduce" : 1,
"output" : 1
},
"ok" : 1,
}
Display this temporary collection:
> db.average_util.find()
{ "_id" : "738b4794-d5ec-48d0-b669-d31d368986ed", "value" : 5.473252670153079 }