[
https://issues.apache.org/jira/browse/HDFS-16845?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17636802#comment-17636802
]
ASF GitHub Bot commented on HDFS-16845:
---------------------------------------
mkuchenbecker commented on code in PR #5142:
URL: https://github.com/apache/hadoop/pull/5142#discussion_r1028372645
##########
hadoop-hdfs-project/hadoop-hdfs-rbf/src/test/java/org/apache/hadoop/hdfs/server/federation/router/TestObserverWithRouter.java:
##########
@@ -439,4 +440,60 @@ public void testRouterMsync() throws Exception {
assertEquals("Four calls should be sent to active", 4,
rpcCountForActive);
}
+
+ @Test
+ public void testSingleRead() throws Exception {
+ List<? extends FederationNamenodeContext> namenodes = routerContext
+ .getRouter().getNamenodeResolver()
+ .getNamenodesForNameserviceId(cluster.getNameservices().get(0), true);
+ assertEquals("First namenode should be observer",
namenodes.get(0).getState(),
+ FederationNamenodeServiceState.OBSERVER);
+ Path path = new Path("/");
+
+ long rpcCountForActive;
+ long rpcCountForObserver;
+
+ // Send read request
+ fileSystem.listFiles(path, false);
+ fileSystem.close();
+
+ rpcCountForActive = routerContext.getRouter().getRpcServer()
+ .getRPCMetrics().getActiveProxyOps();
+ // getListingCall sent to active.
+ assertEquals("Only one call should be sent to active", 1,
rpcCountForActive);
+
+ rpcCountForObserver = routerContext.getRouter().getRpcServer()
+ .getRPCMetrics().getObserverProxyOps();
+ // getList call should be sent to observer
+ assertEquals("No calls should be sent to observer", 0,
rpcCountForObserver);
+ }
+
+ @Test
+ public void testSingleReadUsingObserverReadProxyProvider() throws Exception {
+ fileSystem.close();
+ fileSystem = routerContext.getFileSystemWithObserverReadProxyProvider();
Review Comment:
This seems wrong to special-case in this way. Either manage it during setup
or set it for every function, but I'd advise against mixing the two.
##########
hadoop-hdfs-project/hadoop-hdfs-rbf/src/test/java/org/apache/hadoop/hdfs/server/federation/router/TestObserverWithRouter.java:
##########
@@ -122,7 +123,9 @@ public void startUpCluster(int numberOfObserver,
Configuration confOverrides) th
cluster.waitActiveNamespaces();
routerContext = cluster.getRandomRouter();
- fileSystem = routerContext.getFileSystemWithObserverReadsEnabled();
+ Configuration confToEnableObserverRead = new Configuration();
+
confToEnableObserverRead.setBoolean(HdfsClientConfigKeys.DFS_RBF_OBSERVER_READ_ENABLE,
true);
+ fileSystem = routerContext.getFileSystem(confToEnableObserverRead);
Review Comment:
We are losing coverage on `getFileSystemWithObserverReadsEnabled` with this
change; we should likely be testing both as they are both valid use-cases
whether you want to msync or not.
##########
hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/NameNodeProxiesClient.java:
##########
@@ -349,6 +349,13 @@ public static ClientProtocol
createProxyWithAlignmentContext(
boolean withRetries, AtomicBoolean fallbackToSimpleAuth,
AlignmentContext alignmentContext)
throws IOException {
+ if (conf.getBoolean(HdfsClientConfigKeys.DFS_RBF_OBSERVER_READ_ENABLE,
Review Comment:
What was the original behaviour where someone passed in `null` to this
function?
> Add configuration flag to enable observer reads on routers without using
> ObserverReadProxyProvider
> --------------------------------------------------------------------------------------------------
>
> Key: HDFS-16845
> URL: https://issues.apache.org/jira/browse/HDFS-16845
> Project: Hadoop HDFS
> Issue Type: Bug
> Reporter: Simbarashe Dzinamarira
> Assignee: Simbarashe Dzinamarira
> Priority: Critical
> Labels: pull-request-available
>
> In order for clients to have routers forward their reads to observers, the
> clients must use a proxy with an alignment context. This is currently
> achieved by using the ObserverReadProxyProvider.
> Using ObserverReadProxyProvider allows backward compatible for client
> configurations.
> However, the ObserverReadProxyProvider forces an msync on initialization
> which is not required with routers.
> Performing msync calls is more expensive with routers because the router fans
> out the cal to all namespaces, so we'd like to avoid this.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
