[
https://issues.apache.org/jira/browse/GEODE-10243?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17526012#comment-17526012
]
ASF subversion and git services commented on GEODE-10243:
---------------------------------------------------------
Commit 740d5e56758566146a1034f885197b742832c436 in geode's branch
refs/heads/develop from Jinmei Liao
[ https://gitbox.apache.org/repos/asf?p=geode.git;h=740d5e5675 ]
GEODE-10243: Fail early if old client auth expires (#7603)
* change the default "waitForReAuth" time to 60 seconds
> Old clients with durable queues should fail early if
> AuthenticationExpiredException is thrown
> ---------------------------------------------------------------------------------------------
>
> Key: GEODE-10243
> URL: https://issues.apache.org/jira/browse/GEODE-10243
> Project: Geode
> Issue Type: Improvement
> Components: client queues
> Reporter: Dan Smith
> Assignee: Dan Smith
> Priority: Major
> Labels: pull-request-available
>
> As part of the changes for GEODE-9457, when an AuthenticationExpiredException
> is thrown from the SecurityManager during message dispatching, we send a
> message to 1.15 and newer clients asking them to re-authenticate.
> For 1.14 and older clients, we do not send a message. Instead, we just wait
> for the {color:#00875a}reauthenticate.wait.time{color} to elapse and then
> close the connection.
> The net effect of this is that if users are doing cache operations from 1.14
> and older clients, and their SecurityManager expires the credentials of the
> old clients, they will sometimes see their clients re-authenticate themselves
> in that time window. This will mislead users into thinking that
> re-authentication works with old clients and client queues, even though we
> [have documented that we don't support
> it|https://github.com/apache/geode/blob/09b8b46ef2fa1d463be885c6fa39dbfe1f0e3e83/geode-docs/managing/security/implementing_authentication_expiry.html.md.erb#L35].
> Instead of allowing re-authentication to sometimes work in this unsupported
> use case, we should always fail so that is clear to users that this use case
> is not supported.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
