[
https://issues.apache.org/jira/browse/PHOENIX-6681?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17521299#comment-17521299
]
ASF GitHub Bot commented on PHOENIX-6681:
-----------------------------------------
gokceni commented on code in PR #1413:
URL: https://github.com/apache/phoenix/pull/1413#discussion_r848780886
##########
phoenix-core/src/main/java/org/apache/phoenix/index/IndexMaintainer.java:
##########
@@ -232,7 +232,15 @@ public static void serialize(PTable dataTable,
ImmutableBytesWritable ptr,
}
int nIndexes = indexes.size();
if (dataTable.getTransformingNewTable() != null) {
- nIndexes++;
+ boolean disabled =
dataTable.getTransformingNewTable().getIndexState()!= null &&
+
dataTable.getTransformingNewTable().getIndexState().isDisabled();
+ if (disabled && nIndexes == 0) {
+ ptr.set(ByteUtil.EMPTY_BYTE_ARRAY);
+ return;
+ }
+ if (!disabled) {
+ nIndexes++;
Review Comment:
We are reusing index maintainer code path and TransformMaintainer is the
subclass of IndexMaintainer. So, we are treating it as an index.
##########
phoenix-core/src/main/java/org/apache/phoenix/index/IndexMaintainer.java:
##########
@@ -248,11 +256,15 @@ public static void serialize(PTable dataTable,
ImmutableBytesWritable ptr,
output.write(protoBytes);
}
if (dataTable.getTransformingNewTable() != null) {
- ServerCachingProtos.TransformMaintainer proto =
TransformMaintainer.toProto(
-
dataTable.getTransformingNewTable().getTransformMaintainer(dataTable,
connection));
- byte[] protoBytes = proto.toByteArray();
- WritableUtils.writeVInt(output, protoBytes.length);
- output.write(protoBytes);
+ boolean disabled =
dataTable.getTransformingNewTable().getIndexState()!= null &&
Review Comment:
yes
> Enable new indexes to be optionally created in CREATE_DISABLED state
> ---------------------------------------------------------------------
>
> Key: PHOENIX-6681
> URL: https://issues.apache.org/jira/browse/PHOENIX-6681
> Project: Phoenix
> Issue Type: New Feature
> Reporter: Gokcen Iskender
> Assignee: Gokcen Iskender
> Priority: Major
>
> When we create/drop some indexes in one datacenter, the replication pair
> doesn't immediately have this index and Hbase throws a replication error when
> we try to write into indexes that don't have a matching table on the
> replication pair.
> To remediate this issue, we can optionally create them in CREATE_DISABLED
> state and enable them after all the replication peers have the table.
> Same is true for dropping indexes. When they are dropped we can first disable
> them and drop them later.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
