[
https://issues.apache.org/jira/browse/PHOENIX-5178?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16790068#comment-16790068
]
Ankit Singhal commented on PHOENIX-5178:
----------------------------------------
bq. How did this slip by tests? Something to do with how we run ITs?
Actually, the problem is with a special case and was not exposed until the user
has a mix of user permissions on SYSTEM: and doesn't have all the fixes in
their build. Because in master, an exception handling of AccessDeniedException
is covering this bug and not affecting the user connection as it's timestamp is
always greater than the meta timestamp.
I have updated the patch to just test this specific case by impersonating meta
connection because it will be difficult to reproduce and test the upgrade path
in master which has so many exception handling
> SYSTEM schema is not getting cached at MetaData server
> ------------------------------------------------------
>
> Key: PHOENIX-5178
> URL: https://issues.apache.org/jira/browse/PHOENIX-5178
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.8.0
> Reporter: Ankit Singhal
> Assignee: Ankit Singhal
> Priority: Major
> Attachments: PHOENIX-5178.patch
>
>
> During initialization, the meta connection will not be able to see the SYSTEM
> schema as the scanner at meta server is running with max_timestamp of
> MIN_SYSTEM_TABLE_TIMESTAMP(exclusive) which result in a new connection to
> create SYSTEM schema metadata everytime.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
