[
https://issues.apache.org/jira/browse/KUDU-3223?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17340619#comment-17340619
]
ASF subversion and git services commented on KUDU-3223:
-------------------------------------------------------
Commit a9b8a772bfe4f7be827068b0cf571d8f5f1d362e in kudu's branch
refs/heads/master from Hongjiang Zhang
[ https://gitbox.apache.org/repos/asf?p=kudu.git;h=a9b8a77 ]
KUDU-3223: Management of per-table level limit
When the table's size or row exceeds a threshold, the insert and update
privilege will be removed from the authz token, only scan and delete
privilege are allowed.
Admin is responsible to change the on_disk_size_limit or
row_count_limiti per-table. -1 means no limit.
This feature depends on authz token expiration, and it caused a delay
for write forbidden. Anyway, it provides an approximately management for
write.
Change-Id: I2dbf365ad59f17c0a4e2e7ea6a5afaa7680724b0
Reviewed-on: http://gerrit.cloudera.org:8080/17273
Tested-by: Kudu Jenkins
Reviewed-by: Andrew Wong <aw...@cloudera.com>
> Total table size quota management for every user
> ------------------------------------------------
>
> Key: KUDU-3223
> URL: https://issues.apache.org/jira/browse/KUDU-3223
> Project: Kudu
> Issue Type: Wish
> Reporter: Redriver
> Priority: Major
>
> *Requirement description*
> In production environment, the quota management for total tables size and
> single table size of a single user is necessary. If the quota is reached, the
> write operation including insert and update is forbidden, only table
> creation, scan and delete are allowed.
> We need to allow admin to configure the quota for every user, monitor the
> quota status, and stop the write operation with an error returned to client.
> So, I hope Kudu cli can list the status. Currently, the kudu table statistics
> only gives the disk size and live row count. It is better improve this by
> adding user name for table statistics, or create a new parameter to query
> every table size for every user, and disable/enable a user's write privilege.
> Can this be configured through Apache Ranger?
> *Typical scenarios*
> # Some user has malicious operation to create a huge table, or many small
> tables, as a result, the total table size exceeds the cluster's capacity, we
> have to stop him in time.
> # When writing Kudu table through Flink, the streaming process is always
> running, which means the Kudu table writing never stopped. If we do not have
> a protection, the Kudu cluster will be exhausted someday.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
