[
https://issues.apache.org/jira/browse/PHOENIX-3767?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15958173#comment-15958173
]
Karan Mehta commented on PHOENIX-3767:
--------------------------------------
Since HDFS starts with multiple spans with traceId as 0, the current patch
filters all the spans for which traceId is 0 as they cannot be traced properly.
Relevant HDFS bug is HDFS-11622.
> Prevent spurious tracing spans from Hadoop being written in Phoenix Trace
> tables
> --------------------------------------------------------------------------------
>
> Key: PHOENIX-3767
> URL: https://issues.apache.org/jira/browse/PHOENIX-3767
> Project: Phoenix
> Issue Type: Bug
> Reporter: Karan Mehta
> Assignee: Karan Mehta
>
> PHOENIX-3752 is the JIRA to remove Hadoop Metrics-2 system from the Phoenix
> Framework, whose task is to deliver the received HTrace spans to the phoenix
> table. When replacing this system with a simple one as discussed on this
> JIRA, one of the current test {{testClientServerIndexingTracing()}} triggers
> a bug in Hadoop (HDFS-11583), which causes spurious {{writeTo}} spans being
> generated from {{DataStreamer}} class.
> Since Phoenix implements its {{SpanReceiver}} to receive the HTrace spans,
> all these spans are redirected over here, which eventually get persisted into
> Phoenix tables with the system developed in PHOENIX-3752. The information is
> useless and might use up unnecessary space.
> This JIRA is to add a patch to ignore those spans and prevent them from being
> written to Phoenix tables.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
