virajjasani commented on code in PR #4924:
URL: https://github.com/apache/hbase/pull/4924#discussion_r1046599163
##########
hbase-server/src/main/java/org/apache/hadoop/hbase/replication/regionserver/ReplicationSink.java:
##########
@@ -265,6 +266,11 @@ public void replicateEntries(List<WALEntry> entries, final
CellScanner cells,
mutation.setClusterIds(clusterIds);
mutation.setAttribute(ReplicationUtils.REPLICATION_ATTR_NAME,
HConstants.EMPTY_BYTE_ARRAY);
+ if (attributeList != null) {
Review Comment:
> And if phoenix wants this feature, please describe more about at lease one
usage. Let's see how to better support it.
So let me take a simple example. Phoenix supports multi-tenancy by providing
multi-tenant connection. Hence, same table can be shared by multiple tenants.
This requires tenant id to be a rowkey prefix in hbase rowkey. When a
particular tenant creates tenant connection and writes data, those data are not
visible to other tenants.
Tenant id is very basic entity in phoenix. Now let's say we want to
introduce some level of caching in phoenix server side, the cache would contain
tenant id.
For HBase sink cluster, mutation is already created for specific table
because table name is available in WALKey. For phoenix, let's say table name is
available and mutation has full rowkey as well but it's very difficult to
derive tenant id from this rowkey. Hence, phoenix coproc needs to know of
tenant id attribute with mutation so that it can form tenant level connection
or derive further attributes at sink side and so on.
In addition to tenant id, there are a few more imp metadata that is required
at sink side coproc, without which it's not possible to be aware of how the
data is meant to be used.
> How do you read these attributes at the source cluster's coproc?
If we take same example as above, multi tenant connection is already created
by client and hence tenant id is always available with all operations performed
at source cluster coproc. Besides some of the imp attributes are already
attached as mutation attributes at source side. But when the mutations are
generated at sink side, while hbase metadata is available, coproc like phoenix
metadata is not available and it would be great if we can attach these metadata
attributes to sink mutation, otherwise sink coproc would not be able to derive
them.
Making this configurable feature with proper documentation is also fine, if
you don't mind. WDYT?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
