[jira] [Commented] (YARN-9139) Simplify initializer code of GpuDiscoverer

["Peter Bacsko (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/YARN-9139?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16748080#comment-16748080 ] 

Peter Bacsko commented on YARN-9139:
------------------------------------

[~snemeth] please update the logic with the following: after setting {{pathOfGpuBinary}}, validate that this file actually exists (and executable). It can have a bogus value if this code path runs:
{noformat}
	    } else if (binaryPath.isDirectory()) {
	      binaryPath = new File(binaryPath, DEFAULT_BINARY_NAME);
	      LOG.warn("Specified path is a directory, use " + DEFAULT_BINARY_NAME
	          + " under the directory, updated path-to-executable:"
	          + binaryPath.getAbsolutePath());
	    }
			
	    pathOfGpuBinary = binaryPath.getAbsolutePath();
{noformat}

What if there is no {{nvidia-smi}} under {{binaryPath}}? We must check that and fail immediately and don't wait until {{Shell.execCommand()}} fails later with a probably more cryptic error message. 

> Simplify initializer code of GpuDiscoverer
> ------------------------------------------
>
>                 Key: YARN-9139
>                 URL: https://issues.apache.org/jira/browse/YARN-9139
>             Project: Hadoop YARN
>          Issue Type: Improvement
>            Reporter: Szilard Nemeth
>            Assignee: Szilard Nemeth
>            Priority: Major
>         Attachments: YARN-9139.001.patch
>
>




--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org