[
https://issues.apache.org/jira/browse/YARN-8096?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16420799#comment-16420799
]
Íñigo Goiri commented on YARN-8096:
-----------------------------------
[~oshevchenko], you can make the UPDATE_INTERVAL configurable to a small value
and then check that in the beginning there is an update, then it doesn't update
and after the UPDATE_INTERVAL passes, there is another update. I would use
GenericTestUtils for waiting for these events to happens instead of purely
hardcoded sleeps.
> Wrong condition in AmIpFilter#getProxyAddresses() to update the proxy IP list
> -----------------------------------------------------------------------------
>
> Key: YARN-8096
> URL: https://issues.apache.org/jira/browse/YARN-8096
> Project: Hadoop YARN
> Issue Type: Bug
> Affects Versions: 3.0.0
> Reporter: Oleksandr Shevchenko
> Assignee: Oleksandr Shevchenko
> Priority: Major
> Attachments: YARN-8096.001.patch
>
>
> In AmIpFilter#getProxyAddresses() we have the following condition:
> {code:java}
> if (proxyAddresses == null || (lastUpdate + UPDATE_INTERVAL) >= now) {
> //update RM address
> }
> {code}
> By design, the address should be updated if the last update was more then 5
> min ago. But as we see this condition is wrong.
> Currently, RM address updates permanently. But after 5 minutes after the last
> update, RM address will never be updated again. As a result, we are always
> redirected to the fail page that was added by YARN-4767, even if a network
> issue is resolved now.
> So, we should change this condition to:
> {code:java}
> if (proxyAddresses == null || (lastUpdate + UPDATE_INTERVAL) <= now) \{
> //update RM address
> }
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
