[ https://issues.apache.org/jira/browse/HDFS-8820?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14641240#comment-14641240 ]
Arpit Agarwal commented on HDFS-8820: ------------------------------------- Thanks [~mingma]. bq. Are you going to enable both fair call queue and rpc backoff? No I don't intend to enable fair call queue by default for now. bq. Optionally we might want to adjust dfs.client.retry.max.attempts setting so that it can retry more in the case of NN congestion. Interesting, thanks for the hint. I agree a maximum of 90 seconds may be a little short but not sure what is the right default for a range of deployments. We could file a separate Jira to change this default although I probably won't be able to test out alternatives myself. If anyone else is curious about the math: initial backoff = 500ms, maximum backoff = 15s. sum(.5, 1, 2, 4, 8, 15 x 5) ~ 90 seconds. Initially we'll change the default for just HDFS so I moved this under HDFS. > Enable RPC Congestion control by default > ---------------------------------------- > > Key: HDFS-8820 > URL: https://issues.apache.org/jira/browse/HDFS-8820 > Project: Hadoop HDFS > Issue Type: Improvement > Components: namenode > Reporter: Arpit Agarwal > Assignee: Arpit Agarwal > > We propose enabling RPC congestion control introduced by HADOOP-10597 by > default. > We enabled it on a couple of large clusters a few weeks ago and it has helped > keep the namenodes responsive under load. -- This message was sent by Atlassian JIRA (v6.3.4#6332)