[
https://issues.apache.org/jira/browse/HTRACE-222?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14646580#comment-14646580
]
Colin Patrick McCabe commented on HTRACE-222:
---------------------------------------------
I thought about this a little bit more. I think having {{SpanReceivers}} be
per-Tracer might be more trouble than its worth. I can't think of a case where
we'd want to have a different set of {{SpanReceivers}} active for each Tracer--
but I can think of many cases where this would create the potential for
misconfiguration.
I think that the list of active {{SpanReceivers}} should be global (like it is
currently), but we should close all of them when the last {{Tracer}} is closed,
or when the process is exiting. So instead of reference-counting each
{{SpanReceiver}}, we simply keep a count of the number of active {{Tracer}}
objects. I also think we should never create more than one copy of each type
of {{SpanReceiver}}.
Another thing to keep in mind is that Hadoop has an API for adding and removing
{{SpanReceivers}} at runtime. So we need to support that. To make that API
really useful, we probably also need an API for adding {{Samplers}} to
individual {{Tracer}} objects at runtime as well.
> Add SpanReceiverPool
> --------------------
>
> Key: HTRACE-222
> URL: https://issues.apache.org/jira/browse/HTRACE-222
> Project: HTrace
> Issue Type: Sub-task
> Affects Versions: 4.0
> Reporter: Colin Patrick McCabe
> Assignee: Colin Patrick McCabe
> Attachments: HTRACE-222.001.patch
>
>
> Add SpanReceiverPool, a pool of span receivers which can be used by multiple
> Tracer objects.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
