Jump to: navigation, search

Difference between revisions of "Zaqar/bp/ops-stats"

(Stats Collection)
m (Malini moved page Marconi/bp/ops-stats to Zaqar/bp/ops-stats: Project Rename)
 
(3 intermediate revisions by one other user not shown)
Line 1: Line 1:
== Stats Collection ==
+
== Stats Collector ==
  
We need to gather several stats in MongoDB or with ceilometer. Some of these will need to be exposed via the API, but the first step is recording, which is what this bp is all about.
+
We need to gather several stats. Some of these will need to be exposed via the API (bp/advanced-queue-stats), but the first step is recording, which is what this bp is all about.
  
 
Note: Use a storage driver middleware approach for metering/stats initially (ala Swift)
 
Note: Use a storage driver middleware approach for metering/stats initially (ala Swift)
  
 
Here are some possible metrics (starred items are most important).
 
Here are some possible metrics (starred items are most important).
 +
 +
== Per Cluster ===
 +
* Storage stats
 +
* Storage health
  
 
=== Per account ===
 
=== Per account ===

Latest revision as of 18:42, 7 August 2014

Stats Collector

We need to gather several stats. Some of these will need to be exposed via the API (bp/advanced-queue-stats), but the first step is recording, which is what this bp is all about.

Note: Use a storage driver middleware approach for metering/stats initially (ala Swift)

Here are some possible metrics (starred items are most important).

Per Cluster =

  • Storage stats
  • Storage health

Per account

  • Number of queues*
  • Total messages/day
  • Average messages/sec*

Per queue

  • Average messages/sec enqueued*
  • Number of messages*
  • Life of queue (how long since created)
  • Number of producers
  • Number of workers
  • Number of subscribers
  • Average message size
  • Average message TTL
  • Average message lifetime (explicitly deleted)
  • Average claim TTL
  • Average claim lifetime
  • Average number of messages posted per request
  • Average number of messages requested per GET or claim

Operations should be able to aggregate and graph these metrics across acounts.

See also: http://eavesdrop.openstack.org/meetings/marconi/2013/marconi.2013-06-27-19.05.log.html