[
https://issues.apache.org/jira/browse/HDFS-14150?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16729296#comment-16729296
]
Yiqun Lin commented on HDFS-14150:
----------------------------------
Thanks for updating the initial patch, [~tasanuma0829].
{quote}
When the mount point has one sub-cluster, the quota definition in State Store
is updated using the sub-cluster's quota definition.
{quote}
I'm still thinking for this we need to sync the quota from subcluster to mount
table rather than mount table quota definition to subcluster. From my original
design, I was planning to sync the subcluster‘s quota from State Store
periodically. And making subcluster‘s quota always be consistent with mount
table. This is based on an assumption that once users enable the Router Quota,
the router quota has priority over the subcluster's quota.
> RBF: Quotas of the sub-cluster should be removed when removing the mount point
> ------------------------------------------------------------------------------
>
> Key: HDFS-14150
> URL: https://issues.apache.org/jira/browse/HDFS-14150
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: Takanobu Asanuma
> Assignee: Takanobu Asanuma
> Priority: Major
> Labels: RBF
> Attachments: HDFS-14150.1.patch
>
>
> From HDFS-14143
> {noformat}
> $ hdfs dfsrouteradmin -add /ns1_data ns1 /data
> $ hdfs dfsrouteradmin -setQuota /ns1_data -nsQuota 10 -ssQuota 10
> $ hdfs dfsrouteradmin -ls /ns1_data
> Source Destinations Owner
> Group Mode Quota/Usage
> /ns1_data ns1->/data tasanuma
> users rwxr-xr-x [NsQuota: 10/1, SsQuota:
> 10 B/0 B]
> $ hdfs dfsrouteradmin -rm /ns1_data
> $ hdfs dfsrouteradmin -add /ns1_data ns1 /data
> $ hdfs dfsrouteradmin -ls /ns1_data
> Source Destinations Owner
> Group Mode Quota/Usage
> /ns1_data ns1->/data tasanuma
> users rwxr-xr-x [NsQuota: -/-, SsQuota:
> -/-]
> $ hadoop fs -put file1 /ns1_data/file1
> put: The DiskSpace quota of /data is exceeded: quota = 10 B = 10 B but
> diskspace consumed = 402653184 B = 384 MB
> {noformat}
> This is because the quotas of the subclusters still remain after "hdfs
> dfsrouteradmin -rm". And "hdfs dfsrouteradmin -add" doesn't reflect the
> existing quotas.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
