You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-issues@hadoop.apache.org by "Steve Loughran (Jira)" <ji...@apache.org> on 2020/03/11 20:30:00 UTC

[jira] [Commented] (HADOOP-16819) Possible inconsistent state of AbstractDelegationTokenSecretManager

    [ https://issues.apache.org/jira/browse/HADOOP-16819?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17057364#comment-17057364 ] 

Steve Loughran commented on HADOOP-16819:
-----------------------------------------

# still waiting for that github PR
# as this goes near code I don't understand, we will need to chase down other reviewers; [~omalley] springs to mind

> Possible inconsistent state of AbstractDelegationTokenSecretManager
> -------------------------------------------------------------------
>
>                 Key: HADOOP-16819
>                 URL: https://issues.apache.org/jira/browse/HADOOP-16819
>             Project: Hadoop Common
>          Issue Type: Sub-task
>          Components: fs/s3, security
>    Affects Versions: 3.3.0
>            Reporter: Hankó Gergely
>            Assignee: Hankó Gergely
>            Priority: Major
>         Attachments: HADOOP-16819.001.patch
>
>
> [AbstractDelegationTokenSecretManager.updateCurrentKey|https://github.com/apache/hadoop/blob/581072a8f04f7568d3560f105fd1988d3acc9e54/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/security/token/delegation/AbstractDelegationTokenSecretManager.java#L360] increments the current key id and creates the new delegation key in two distinct synchronized blocks.
> This means that other threads can see the class in an *inconsistent state, where the key for the current key id doesn't exist (yet)*.
> For example the following method sometimes returns null when the token remover thread is between the two synchronized blocks:
> {noformat}
> @Override
> public DelegationKey getCurrentKey() {
>   return getDelegationKey(getCurrentKeyId());
> }{noformat}
>  
> Also it is possible that updateCurrentKey is called from multiple threads at the same time so *distinct keys can be generated with the same key id*.
>  
> This issue is suspected to be the cause of the intermittent failure of  [TestLlapSignerImpl.testSigning|https://github.com/apache/hive/blob/3c0705eaf5121c7b61f2dbe9db9545c3926f26f1/llap-server/src/test/org/apache/hadoop/hive/llap/security/TestLlapSignerImpl.java#L195] - HIVE-22621.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

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