[
https://issues.apache.org/jira/browse/YARN-9768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16938819#comment-16938819
]
Iñigo Goiri commented on YARN-9768:
-----------------------------------
Thanks [[email protected]] for [^YARN-9768.002.patch].
As we are using getTimeDuration(), the variables should be also time durations,
I usually do:
{code}
public static final long
DEFAULT_RM_DELEGATION_TOKEN_RENEWER_THREAD_RETRY_INTERVAL =
TimeUni.SECONDS.toMillis(60);
{code}
Regarding reading these variables, I prefer using the following indentation:
{code}
tokenRenewerThreadRetryInterval = conf.getTimeDuration(
YarnConfiguration.RM_DELEGATION_TOKEN_RENEWER_THREAD_RETRY_INTERVAL,
YarnConfiguration.DEFAULT_RM_DELEGATION_TOKEN_RENEWER_THREAD_RETRY_INTERVAL,
TimeUnit.MILLISECONDS);
{code}
DelegationTokenRenewer#215 should be a single line.
In DelegationTokenRenewer#227, you should do {{catch(TimeOutException toe)}}
then add an extra {{catch(Exception e)}}.
I also think DelegationTokenRenewer#234 can be a lambda.
Avoid DelegationTokenRenewer#442, it just adds churn in an unrelated patch.
Same for #691 and #508.
Why are we making DelegationTokenRenewer#551 a debug message? If we change
that, let's also use logger style with {}.
DelegationTokenRenewer#1107 should be a single line. Same as 1129 and 1047 with
the end of file.
For TestDelegationTokenRenewer, let's also avoid the changes like #169.
#1567 should be a single line. Same for 1571 and 1573.
I'm not a big fan of this long sleeps (#1591).
You have a print in 1593, which could be done properly adding a message to the
assertTrue (which could use a extracted version of the conf).
> RM Renew Delegation token thread should timeout and retry
> ---------------------------------------------------------
>
> Key: YARN-9768
> URL: https://issues.apache.org/jira/browse/YARN-9768
> Project: Hadoop YARN
> Issue Type: Improvement
> Reporter: CR Hota
> Priority: Major
> Attachments: YARN-9768.001.patch, YARN-9768.002.patch
>
>
> Delegation token renewer thread in RM (DelegationTokenRenewer.java) renews
> HDFS tokens received to check for validity and expiration time.
> This call is made to an underlying HDFS NN or Router Node (which has exact
> APIs as HDFS NN). If one of the nodes is bad and the renew call is stuck the
> thread remains stuck indefinitely. The thread should ideally timeout the
> renewToken and retry from the client's perspective.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
