[
https://issues.apache.org/jira/browse/HBASE-26910?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Bryan Beaudreault updated HBASE-26910:
--------------------------------------
Description:
HBASE-20333 added a new shaded artifact, hbase-shaded-client-byo-hadoop. This
is useful in environments where they already depend on hadoop so pulling in the
hadoop classes from hbase-shaded-client would be extreme.
I think the split that [~busbey] decided on there definitely made sense for the
general case (with byo-hadoop being the specially named artifact). But some of
us may work in environments where the default case is the byo-hadoop case. For
those environments, it's a little confusing and verbose to have to refer to the
hbase client as hbase-shaded-client-byo-hadoop everywhere. For example, we have
well over 500 poms directly pulling in hbase client and many more pulling it in
transitively.
Using profiles I think we could provide an optional activation which allows
someone building their own hbase to decide whether hbase-shaded-client should
include hadoop or not. This way normal users using the artifacts published to
maven central are unaffected, but super users building their own have some
extra flexibility.
was:
HBASE-20333 added a new shaded artifact, hbase-shaded-client-byo-hadoop. This
is useful in environments where they already depend on hadoop so pulling in the
hadoop classes from hbase-shaded-client would be extreme.
I think the split that [~busbey] decided on there definitely made sense for the
general case (with byo-hadoop being the specially named artifact). But some of
us may work in environments where the default case is the byo-hadoop case. For
those environments, it's a little confusing and verbose to have to refer to the
hbase client as hbase-shaded-client-byo-hadoop everywhere.
Using profiles I think we could provide an optional activation which allows
someone building their own hbase to decide whether hbase-shaded-client should
include hadoop or not. This way normal users using the artifacts published to
maven central are unaffected, but super users building their own have some
extra flexibility.
> Allow builders of hbase to specify hadoop inclusion mode for
> hbase-shaded-client
> --------------------------------------------------------------------------------
>
> Key: HBASE-26910
> URL: https://issues.apache.org/jira/browse/HBASE-26910
> Project: HBase
> Issue Type: Improvement
> Reporter: Bryan Beaudreault
> Assignee: Bryan Beaudreault
> Priority: Major
>
> HBASE-20333 added a new shaded artifact, hbase-shaded-client-byo-hadoop. This
> is useful in environments where they already depend on hadoop so pulling in
> the hadoop classes from hbase-shaded-client would be extreme.
> I think the split that [~busbey] decided on there definitely made sense for
> the general case (with byo-hadoop being the specially named artifact). But
> some of us may work in environments where the default case is the byo-hadoop
> case. For those environments, it's a little confusing and verbose to have to
> refer to the hbase client as hbase-shaded-client-byo-hadoop everywhere. For
> example, we have well over 500 poms directly pulling in hbase client and many
> more pulling it in transitively.
> Using profiles I think we could provide an optional activation which allows
> someone building their own hbase to decide whether hbase-shaded-client should
> include hadoop or not. This way normal users using the artifacts published to
> maven central are unaffected, but super users building their own have some
> extra flexibility.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
