Github user justinleet commented on a diff in the pull request:
https://github.com/apache/incubator-metron/pull/486#discussion_r107534534
--- Diff: pom.xml ---
@@ -67,20 +67,44 @@
<!-- base project versions -->
<base_storm_version>1.0.1</base_storm_version>
<base_flux_version>1.0.1</base_flux_version>
- <base_kafka_version>0.10.0.1</base_kafka_version>
+ <base_kafka_version>0.10.0</base_kafka_version>
<base_hadoop_version>2.7.1</base_hadoop_version>
<base_hbase_version>1.1.1</base_hbase_version>
- <global_accumulo_version>1.8.0</global_accumulo_version>
<base_flume_version>1.5.2</base_flume_version>
<!-- full dependency versions -->
+ <global_accumulo_version>1.8.0</global_accumulo_version>
<global_antlr_version>4.5</global_antlr_version>
<global_opencsv_version>3.7</global_opencsv_version>
<global_curator_version>2.7.1</global_curator_version>
<global_classindex_version>3.3</global_classindex_version>
- <global_storm_version>${base_storm_version}</global_storm_version>
+ <global_storm_version>1.0.3</global_storm_version>
+ <!--
+ This bears some explanation. storm-kafka-client is our kafka
spout.
+ If we ever hope to support kerberos, this provides the
capability to do so
+ in apache. Unfortunately, it also does not support, as of
Storm 1.0.x
+ Kafka 0.10.x (see
https://issues.apache.org/jira/browse/STORM-2091).
+ The consumer libraries (not to be confused with the protocol)
on the JVM
+ are binary incompatible. Note the discussion on
https://issues.apache.org/jira/browse/KAFKA-3006,
+ the main issue is the move to Collection over List. While
this would be polymorphically
+ a non-issue, it would require a recompile of
storm-kafka-client against Kafka 0.10.x.
+
+ Since a targeted platform is HDP 2.5.x, which ships only
kafka 0.10.x, we need
+ to support kafka 0.10.x. Therefore, if we are to use this,
then we would need
+ to support both Kafka 0.9.x and 0.10.x. Unfortunately, this
would require us
+ to fork some of the internal projects because the 0.9.x API
has shifted
+ (e.g. the Admin functions have different parameters) and
behaves
+ differently than 0.10.x in subtle ways (e.g. KAFKA_GET
doesn't work as implemented).
+
+ Rather than do this, we chose to depend on the HDP version of
storm-kafka because
+ it is compiled against 0.10.x and therefore would allow us to
not fork our support
+ for kafka. I do not like this bleeding of the HDP profile
dependency into the default,
+ but I justify it by noting that this should be able to be
removed when we migrate to
+ Storm 1.1.x, which properly supports Kafka 0.10.x.
+ -->
+
<global_storm_kafka_version>1.0.1.2.5.0.0-1245</global_storm_kafka_version>
--- End diff --
I'm sure you've already thought of this, but assuming we do go with this,
please make sure this gets a JIRA associated with it.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---