hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Daryn Sharp (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-11246) FSNameSystem#logAuditEvent should be called outside the read or write locks
Date Wed, 19 Jul 2017 21:46:01 GMT

    [ https://issues.apache.org/jira/browse/HDFS-11246?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16093843#comment-16093843

Daryn Sharp commented on HDFS-11246:

I'm not sure I like either of the the suggestions because it re-orders the audit logging and
edit log syncing.  On a semantic level, one might expect the audit log to only include durable
edits.  It may be confusing to see audits for edits lost in a crash.

More importantly logging before syncing is likely to increase the odds of less txns batched
per sync.  Log4j synchronization can be a real bottleneck, even with the async appender support
I added.

The double try feels a bit odd but it seems like a clean change that doesn't cause edit/audit

> FSNameSystem#logAuditEvent should be called outside the read or write locks
> ---------------------------------------------------------------------------
>                 Key: HDFS-11246
>                 URL: https://issues.apache.org/jira/browse/HDFS-11246
>             Project: Hadoop HDFS
>          Issue Type: Bug
>    Affects Versions: 2.7.3
>            Reporter: Kuhu Shukla
>            Assignee: Kuhu Shukla
>         Attachments: HDFS-11246.001.patch, HDFS-11246.002.patch
> {code}
>     readLock();
>     boolean success = true;
>     ContentSummary cs;
>     try {
>       checkOperation(OperationCategory.READ);
>       cs = FSDirStatAndListingOp.getContentSummary(dir, src);
>     } catch (AccessControlException ace) {
>       success = false;
>       logAuditEvent(success, operationName, src);
>       throw ace;
>     } finally {
>       readUnlock(operationName);
>     }
> {code}
> It would be nice to have audit logging outside the lock esp. in scenarios where applications
hammer a given operation several times. 

This message was sent by Atlassian JIRA

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

View raw message