kbarber

lein test :only com.puppetlabs.puppetdb.test.http.facts/factset-queries
FAIL in (factset-queries) (facts.clj:892)
Testing version :v4 factset queries should return appropriate results
expected: (= (into {} (first responses)) {"facts" {"my_structured_fact" {"a" 1, "b" 3.14, "c" ["a" "b" "c"], "d" {"n" ""}, "e" "1", "f" nil}, "domain" "testing.com", "uptime_seconds" "4000", "test#~delimiter" "foo"}, "timestamp" (to-string current-time), "environment" "DEV", "certname" "foo1"})
  actual: (not (= {"facts" {"test#~delimiter" "foo", "uptime_seconds" "4000", "domain" "testing.com", "my_structured_fact" {"b" 3.14, "a" 1, "e" "1", "d" {"n" ""}, "f" nil, "c" ["a" "b" "c"]}}, "timestamp" "2014-07-31T23:27:45.298Z", "environment" "DEV", "certname" "foo1"} {"facts" {"uptime_seconds" "4000", "domain" "testing.com", "test#~delimiter" "foo", "my_structured_fact" {"d" {"n" ""}, "f" nil, "e" "1", "a" 1, "b" 3.14, "c" ["a" "b" "c"]}}, "timestamp" "2014-07-31T23:27:45.297Z", "environment" "DEV", "certname" "foo1"}))
lein test :o

kylog

#!/usr/bin/env ruby

require 'jira'
require 'pp'
require 'getoptlong'

# default values
username = "gepetto-bot"
password = ENV["GEPETTO_BOT_PASSWORD"]

kbarber

JAVA_ARGS="-server -Xmx1024m -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/var/log/puppetdb/puppetdb-oom.hprof  -XX:+UseParNewGC -XX:+UseConcMarkSweepGC -XX:+PrintGCDateStamps -XX:+PrintGCDetails -verbose:gc -Xloggc:/var/log/puppetdb/gc.log"

kbarber

Main reference for tuning:

http://wiki.postgresql.org/wiki/Tuning_Your_PostgreSQL_Server

Most importantly you should make sure you have a way of measuring performance first, and then when you try new changes try them one at a time, gather data (3-5 samples are a good start, you might need to run tests a few times to get take into account cache priming) then compare with old settings.

The most important principle is, the less disk IO the better. So if we have the correct indexing, it works less to find an answer. If you can cache data into RAM, that even better, as the disk IO will be less frequent.

So focus on:

kbarber

diff --git a/indirector/catalog/puppetdb.rb b/indirector/catalog/puppetdb.rb
index 227eeab..6eb6ed0 100644
--- a/indirector/catalog/puppetdb.rb
+++ b/indirector/catalog/puppetdb.rb
@@ -7,9 +7,15 @@ class Puppet::Resource::Catalog::Puppetdb < Puppet::Indirector::REST
   include Puppet::Util::Puppetdb::CommandNames

   def save(request)
+    start = Time.now
+

kbarber

# Generated by iptables-save v1.4.7 on Sun Feb 24 00:48:11 2013
*mangle
:PREROUTING ACCEPT [756:54757]
:INPUT ACCEPT [756:54757]
:FORWARD ACCEPT [0:0]
:OUTPUT ACCEPT [578:527896]
:POSTROUTING ACCEPT [578:527896]
-A INPUT -s 1.1.1.2/32 -j CHECKSUM --checksum-fill 
-A OUTPUT -s 1.1.1.2/32 -j CLASSIFY --set-class 0004:0056 
COMMIT

kbarber

#!/usr/bin/env ruby

require 'yaml'

a = { 
  "a" => "b",
  "e" => "b",
  "c" => "b",
  "p" => "b",
  "z" => "b",

kbarber

#!/usr/bin/env ruby

require 'yaml'

a = { 
  "a" => "b",
  "e" => "b",
  "c" => "b",
  "p" => "b",
  "z" => "b",