[
https://issues.apache.org/jira/browse/HTRACE-237?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14945940#comment-14945940
]
stack commented on HTRACE-237:
------------------------------
So, updating htrace, a 4.0.1 won't be able to write a 4.1? Can we not do both?
bq. It's a common pattern in Hadoop where we want to cache things pulled out of
a Configuration object.
Since when is it good to replicate an hadoop pattern, especially when we don't
have to (smile).
bq, One reason we do it is because accessing Configuration is slow (it's much
slower than a normal hash table lookup) and we want to have configuration
values in an easy-to-access form.
But we are going against HTraceConfiguration Interface, not Hadoop
Configuration? We can do whatever we want int he Implementation? Cache?
bq. If htraced.receiver.packed is false, we use JSON over HTTP. If
htraced.receiver.packed is true, we use msgpack over HRPC.
Ok. Its doc'd? Release note at least.
> Optimize htraced span receiver
> ------------------------------
>
> Key: HTRACE-237
> URL: https://issues.apache.org/jira/browse/HTRACE-237
> Project: HTrace
> Issue Type: Improvement
> Affects Versions: 4.1
> Reporter: Colin Patrick McCabe
> Assignee: Colin Patrick McCabe
> Attachments: HTRACE-237.001.patch, HTRACE-237.002.patch,
> HTRACE-237.003.patch, HTRACE-237.004.patch
>
>
> Optimize htraced span receiver so that it supports double buffering and both
> the msgpack and REST transports.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
