[
https://issues.apache.org/jira/browse/STORM-1700?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15321491#comment-15321491
]
ASF GitHub Bot commented on STORM-1700:
---------------------------------------
Github user HeartSaVioR commented on a diff in the pull request:
https://github.com/apache/storm/pull/1325#discussion_r66345117
--- Diff: conf/storm.yaml.example ---
@@ -39,10 +39,13 @@
# - "server2"
## Metrics Consumers
+## NOTE: task queue will be unbounded when max.retain.metric.tuples is
equal or less than 0.
# topology.metrics.consumer.register:
# - class: "org.apache.storm.metric.LoggingMetricsConsumer"
+# max.retain.metric.tuples: 100
# parallelism.hint: 1
# - class: "org.mycompany.MyMetricsConsumer"
+# max.retain.metric.tuples: 100
--- End diff --
@ptgoetz Good. I'll address according to how we decide to make it pluggable
vs flexible.
> Introduce 'whitelist' / 'blacklist' option to MetricsConsumer
> -------------------------------------------------------------
>
> Key: STORM-1700
> URL: https://issues.apache.org/jira/browse/STORM-1700
> Project: Apache Storm
> Issue Type: Sub-task
> Components: storm-core
> Affects Versions: 1.0.0, 2.0.0
> Reporter: Jungtaek Lim
> Assignee: Jungtaek Lim
>
> Storm provides various metrics by default, and so on some external modules
> (storm-kafka).
> When we register MetricsConsumer, MetricsConsumer should handle all of
> metrics. If MetricsConsumer cannot keep up with these metrics, only way to
> keep up is increasing parallelism, which seems limited. Furthermore, some
> users don't want to care about some metrics since unintended metrics will
> fill external storage.
> Though MetricsConsumer itself can filter metrics by name, it would be better
> to support filter by Storm side. It will reduce the redundant works for Storm
> community.
> If we provide filter options, it would be great.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
