[
https://issues.apache.org/jira/browse/PHOENIX-7251?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17824217#comment-17824217
]
ASF GitHub Bot commented on PHOENIX-7251:
-----------------------------------------
shahrs87 commented on code in PR #1845:
URL: https://github.com/apache/phoenix/pull/1845#discussion_r1515288037
##########
phoenix-core/src/it/java/org/apache/phoenix/end2end/ServerMetadataCacheTestImpl.java:
##########
@@ -0,0 +1,66 @@
+package org.apache.phoenix.end2end;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.hbase.ServerName;
+import org.apache.phoenix.cache.ServerMetadataCacheImpl;
+import org.apache.phoenix.hbase.index.util.ImmutableBytesPtr;
+import org.apache.phoenix.util.QueryUtil;
+import org.apache.phoenix.util.SchemaUtil;
+
+import java.sql.Connection;
+import java.sql.SQLException;
+import java.util.HashMap;
+import java.util.Map;
+import java.util.Properties;
+
+/**
+ * Implementation of {@link ServerMetadataCache} for Integration Tests.
+ * Supports keeping more than one instance keyed on the regionserver
ServerName.
+ */
+public class ServerMetadataCacheTestImpl extends ServerMetadataCacheImpl {
Review Comment:
Can you add some more context in the comment something like:
PhoenixRegionServerEndpoint is a region server coproc. There is a 1-1
correspondence between PhoenixRegionServerEndpoint and ServerMetadataCache.
In ITs we can have multiple regionservers per cluster so we need multiple
instances of ServerMetadataCache.
And HighAvailabilityTestingUtility creates 2 clusters so we need to have one
instance of ServerMetadataCache for each regionserver in each cluster.
> Refactor server-side code to support multiple ServerMetadataCache for HA tests
> ------------------------------------------------------------------------------
>
> Key: PHOENIX-7251
> URL: https://issues.apache.org/jira/browse/PHOENIX-7251
> Project: Phoenix
> Issue Type: Sub-task
> Reporter: Palash Chauhan
> Assignee: Palash Chauhan
> Priority: Major
>
> In the metadata caching re-design, `ServerMetadataCache` is required to be a
> singleton in the implementation. This affects tests for the HA use case
> because the coprocessors on the 2 clusters end up using the same
> `ServerMetadataCache`. All tests which execute queries with 1 of the clusters
> unavailable will fail.
> We can refactor the implementation in the following way to support HA test
> cases:
> 1. Create a `ServerMetadataCache` interface and use the current
> implementation as `ServerMetadataCacheImpl` for all other tests. This would
> be a singleton.
> 2. Implement `ServerMetadataCacheHAImpl` with a map of instances keyed on
> config.
> 3. Extend `PhoenixRegionServerEndpoint` and use `ServerMetadataCacheHAImpl`.
> 4. In HA tests, load this new endpoint on the region servers.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
