[jira] [Commented] (HDFS-15350) Set dfs.client.failover.random.order to true as default
[ https://issues.apache.org/jira/browse/HDFS-15350?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17105474#comment-17105474 ] Hudson commented on HDFS-15350: --- SUCCESS: Integrated in Jenkins build Hadoop-trunk-Commit #18237 (See [https://builds.apache.org/job/Hadoop-trunk-Commit/18237/]) HDFS-15350. Set dfs.client.failover.random.order to true as default. (github: rev 928b81a5339a3d91e77b268d825973a0d9efc1ab) * (edit) hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/ha/TestObserverReadProxyProvider.java * (edit) hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml * (edit) hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/client/HdfsClientConfigKeys.java > Set dfs.client.failover.random.order to true as default > --- > > Key: HDFS-15350 > URL: https://issues.apache.org/jira/browse/HDFS-15350 > Project: Hadoop HDFS > Issue Type: Improvement > Components: hdfs-client >Reporter: Takanobu Asanuma >Assignee: Takanobu Asanuma >Priority: Major > Fix For: 3.4.0 > > > {noformat} > Currently, the default value of dfs.client.failover.random.order is > false. If it's true, clients access to NameNodes random order instead > of the configured order which is defined in hdfs-site.xml. > Setting dfs.client.failover.random.order=true is very important for > RBF if there are multiple routers. If it's false, all the clients > point to the same router because routers are always active. > And I think dfs.client.failover.random.order=true would be good manner > for normal HA(two-NameNodes) Cluster too. If it's false and the first > NameNode is standby, clients always access to standby NameNode at > first. > So I'd like to set dfs.client.failover.random.order to true as default > from 3.4. Does anyone have any concerns? > {noformat} > https://lists.apache.org/thread.html/ra79dde30235a1d302ea82120de8829c0aa7d6c0789f4613430610b8a%40%3Chdfs-dev.hadoop.apache.org%3E -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-15350) Set dfs.client.failover.random.order to true as default
[ https://issues.apache.org/jira/browse/HDFS-15350?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17105022#comment-17105022 ] Takanobu Asanuma commented on HDFS-15350: - [~weichiu] In the case of Multiple Standby NameNodes, I think random.order=true is better even more from the point of load balancing. For example, there are nn1,nn2,nn3,nn4. If random.order=false, the worst case is that nn4 is Active NN. In this case, all clients will always access to the Active NN(nn4) after three retries. In the case of Observer NameNodes, with the same logic, I think random.order=true is better for WRITE operations. > Set dfs.client.failover.random.order to true as default > --- > > Key: HDFS-15350 > URL: https://issues.apache.org/jira/browse/HDFS-15350 > Project: Hadoop HDFS > Issue Type: Improvement > Components: hdfs-client >Reporter: Takanobu Asanuma >Assignee: Takanobu Asanuma >Priority: Major > > {noformat} > Currently, the default value of dfs.client.failover.random.order is > false. If it's true, clients access to NameNodes random order instead > of the configured order which is defined in hdfs-site.xml. > Setting dfs.client.failover.random.order=true is very important for > RBF if there are multiple routers. If it's false, all the clients > point to the same router because routers are always active. > And I think dfs.client.failover.random.order=true would be good manner > for normal HA(two-NameNodes) Cluster too. If it's false and the first > NameNode is standby, clients always access to standby NameNode at > first. > So I'd like to set dfs.client.failover.random.order to true as default > from 3.4. Does anyone have any concerns? > {noformat} > https://lists.apache.org/thread.html/ra79dde30235a1d302ea82120de8829c0aa7d6c0789f4613430610b8a%40%3Chdfs-dev.hadoop.apache.org%3E -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-15350) Set dfs.client.failover.random.order to true as default
[ https://issues.apache.org/jira/browse/HDFS-15350?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17104638#comment-17104638 ] Wei-Chiu Chuang commented on HDFS-15350: What would this mean when you have multiple standby namenodes, or Observer Nodes? > Set dfs.client.failover.random.order to true as default > --- > > Key: HDFS-15350 > URL: https://issues.apache.org/jira/browse/HDFS-15350 > Project: Hadoop HDFS > Issue Type: Improvement > Components: hdfs-client >Reporter: Takanobu Asanuma >Assignee: Takanobu Asanuma >Priority: Major > > {noformat} > Currently, the default value of dfs.client.failover.random.order is > false. If it's true, clients access to NameNodes random order instead > of the configured order which is defined in hdfs-site.xml. > Setting dfs.client.failover.random.order=true is very important for > RBF if there are multiple routers. If it's false, all the clients > point to the same router because routers are always active. > And I think dfs.client.failover.random.order=true would be good manner > for normal HA(two-NameNodes) Cluster too. If it's false and the first > NameNode is standby, clients always access to standby NameNode at > first. > So I'd like to set dfs.client.failover.random.order to true as default > from 3.4. Does anyone have any concerns? > {noformat} > https://lists.apache.org/thread.html/ra79dde30235a1d302ea82120de8829c0aa7d6c0789f4613430610b8a%40%3Chdfs-dev.hadoop.apache.org%3E -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org