You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@pulsar.apache.org by GitBox <gi...@apache.org> on 2019/04/30 17:39:54 UTC

[GitHub] [pulsar] cdbartholomew opened a new pull request #4183: [Issue 3112][Broker Prometheus Metrics]Fix Duplicate Prometheus TYPE fix and add test

cdbartholomew opened a new pull request #4183: [Issue 3112][Broker Prometheus Metrics]Fix Duplicate Prometheus TYPE fix and add test
URL: https://github.com/apache/pulsar/pull/4183
 
 
   
   Fixes #3112 
   
   ### Motivation
   
   In this pull request, I set out to fix #3112, which is caused by duplicate TYPE statements in the metrics output which leads to parsing of Prometheus metrics to fail in recent versions of Prometheus. Because of this, Prometheus will report the broker target as down.
   
   Since I started looking at this, the type definitions have been removed (#4136) from topics metric output. I think these types are useful in Prometheus and have added them back in. 
   
   While testing this fix in my geo-replicated setup, I found a format in error (missing quote and comma) in the TopicStats.metricWithRemoteCluster method. This pull request includes a fix for that issue.
   
   I have also added a new test to PrometheusMetricsTest.java that fails without these changes but passed with them.
   
   ### Modifications
   
   I added a static HashMap to TopicStats to keep track of the TYPEs that have been output. All writing of the TYPE for topics and namespaces is done with the TopicStats.metricType method. I modified that method to update the HashMap and only print the TYPE out for the first occurrence of the metric name.  I also added a method reset the HashMap, which gets called in NamespaceStatsAggregator.generate. 
   
   ### Verifying this change
   
   This change added tests and can be verified as follows:
     - Added testDuplicateMetricTypeDefinitions which checks for:
          - duplicate TYPE definitions in the Prometheus output
          - validates that no TYPE definition appears after the first metric sample
          - ensures that all metrics have a defined type
   
   I execute the test twice to make sure the resetting of the HashMap of the already seen metric type definitions works correctly. This test passes for me reliably (both occurrences).
   
   I have confirmed using promtool that the metrics output will now parse without error using versions 2.7.1 and 2.9.2 (which is the latest). There are many warnings around missing HELP definitions and metrics using reserved suffixes (ex. _count), but no errors.
   
   In addition, I have patched 2.3.1 with this fix and am currently running it in my cluster. Prometheus (2.7.1) successfully parses the metrics and I am able to see namespace and topic-level metrics.
   
   ### Does this pull request potentially affect one of the following parts:
   
   *If `yes` was chosen, please highlight the changes*
   
     - Dependencies (does it add or upgrade a dependency): (no)
     - The public API: no
     - The schema: no
     - The default values of configurations: no
     - The wire protocol: no
     - The rest endpoints: no
     - The admin cli options: no
     - Anything that affects deployment: no
   
   ### Documentation
   
     - Does this pull request introduce a new feature? no
     - If yes, how is the feature documented? not applicable
     - If a feature is not applicable for documentation, explain why?
     - If a feature is not documented yet in this PR, please create a followup issue for adding the documentation
   

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services