You are viewing a plain text version of this content. The canonical link for it is here.
Posted to hdfs-dev@hadoop.apache.org by "Mingliang Liu (JIRA)" <ji...@apache.org> on 2015/11/06 00:19:27 UTC
[jira] [Created] (HDFS-9387) Parse namenodeUri parameter only once
in NNThroughputBenchmark$OperationStatsBase#verifyOpArgument()
Mingliang Liu created HDFS-9387:
-----------------------------------
Summary: Parse namenodeUri parameter only once in NNThroughputBenchmark$OperationStatsBase#verifyOpArgument()
Key: HDFS-9387
URL: https://issues.apache.org/jira/browse/HDFS-9387
Project: Hadoop HDFS
Issue Type: Bug
Components: namenode
Reporter: Mingliang Liu
Assignee: Mingliang Liu
In {{NNThroughputBenchmark$OperationStatsBase#verifyOpArgument()}}, the {{namenodeUri}} is always parsed from {{-namenode}} argument. This works just fine if the {{-op}} parameter is not {{all}}, as the single benchmark will need to parse the {{namenodeUri}} from args anyway.
When the {{-op}} is {{all}}, namely all sub-benchmark will run, multiple sub-benchmark will call the {{verifyOpArgument()}} method. In this case, the first sub-benchmark reads the {{namenode}} argument and removes it from args. The other sub-benchmarks will thereafter read {{null}} value since the argument is removed. This contradicts the intension of providing {{namenode}} for all sub-benchmarks.
{code:title=current code}
try {
namenodeUri = StringUtils.popOptionWithArgument("-namenode", args);
} catch (IllegalArgumentException iae) {
printUsage();
}
{code}
The fix is to parse the {{namenodeUri}}, which is shared by all sub-benchmarks, from {{-namenode}} argument only once. This follows the convention of parsing other global arguments in {{OperationStatsBase#verifyOpArgument()}}.
{code:title=simple fix}
if (args.indexOf("-namenode") >= 0) {
try {
namenodeUri = StringUtils.popOptionWithArgument("-namenode", args);
} catch (IllegalArgumentException iae) {
printUsage();
}
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)