[
https://issues.apache.org/jira/browse/SENTRY-1928?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16156104#comment-16156104
]
Vamsee Yarlagadda commented on SENTRY-1928:
-------------------------------------------
One thing to keep in mind is that, this problem can occur even if we have a
single HMS and Sentry server. This can happen if there is too much load on HMS
that has all it's threads exhausted because of user's requests and it opens a
backdoor connection to Sentry again for sync. And Sentry can fail to get a
connection to Hive to get all the latest notifications.
If this happens, how do we come out of it?
> HMSFollower should close HMS connections when an error to HMS occurs
> --------------------------------------------------------------------
>
> Key: SENTRY-1928
> URL: https://issues.apache.org/jira/browse/SENTRY-1928
> Project: Sentry
> Issue Type: Bug
> Components: Sentry
> Affects Versions: 2.0.0
> Reporter: Sergio Peña
> Assignee: Sergio Peña
> Priority: Critical
>
> Sometimes the HMSFollower fails getting notifications from a HMS-HA
> environment because of SocketTimeoutExceptions. When this exception happens,
> the HMSFollower does not close such connection and it attempts to get
> notifications from the same HMS service causing a deadlock between HMS sync
> notifications listener and Sentry.
> We should close the current HMS connections and re-connect to another HMS
> service if this SocketTimeoutException happens or any other error with the
> HMS happens.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
